Skip to content

P
PaddleClas

PaddlePaddle / PaddleClas
1 年多 前同步成功

通知 116
Star 4999
Fork 1114
P
PaddleClas
提交 f5168a7a 编写于 作者: S shippingwang
浏览文件
操作

Merge branch 'master' of https://github.com/PaddlePaddle/PaddleClas into add_dali

上级 f03fab0e 9d3f36b7
展开全部 隐藏空白更改
内联 并排
Showing with 2244 addition and 155 deletion
README.md
浏览文件 @ f5168a7a
......@@ -22,7 +22,7 @@
<img src="./docs/images/models/V100_benchmark/v100.fp32.bs1.main_fps_top1_s.jpg" width="700">
</div>
上图对比了一些最新的面向服务器端应用场景的模型，在使用V100，FP32和TensorRT，batch size为1时的预测时间及其准确率，图中准确率82.4%的ResNet50_vd_ssld和83.7%的ResNet101_vd_ssld，是采用PaddleClas提供的SSLD知识蒸馏方案训练的模型。图中相同颜色和符号的点代表同一系列不同规模的模型。不同模型的简介、FLOPS、Parameters以及详细的GPU预测时间(包括不同batchsize的T4卡预测速度)请参考文档教程中的[**模型库章节**](https://paddleclas.readthedocs.io/zh_CN/latest/models/models_intro.html)。
上图对比了一些最新的面向服务器端应用场景的模型，在使用V100，FP32和TensorRT，batch size为1时的预测时间及其准确率，图中准确率83.0%的ResNet50_vd_ssld_v2和83.7%的ResNet101_vd_ssld，是采用PaddleClas提供的SSLD知识蒸馏方案训练的模型，其中v2表示在训练时添加了AutoAugment数据增广策略。图中相同颜色和符号的点代表同一系列不同规模的模型。不同模型的简介、FLOPS、Parameters以及详细的GPU预测时间(包括不同batchsize的T4卡预测速度)请参考文档教程中的[**模型库章节**](https://paddleclas.readthedocs.io/zh_CN/latest/models/models_intro.html)。
<div align="center">
<img
......
README_en.md
浏览文件 @ f5168a7a
......@@ -2,11 +2,9 @@
# PaddleClas
> **English version of PaddleClas. Updating...**
**Book**: https://paddleclas-en.readthedocs.io/en/latest/
**Book**: https://paddleclas.readthedocs.io
**Quick start PaddleClas in 30 minutes**: https://paddleclas.readthedocs.io/zh_CN/latest/tutorials/quick_start.html
**Quick start PaddleClas in 30 minutes**: https://paddleclas-en.readthedocs.io/en/latest/tutorials/quick_start_en.html
## Introduction
......@@ -18,14 +16,14 @@ PaddleClas is a toolset for image classification tasks prepared for the industry
## Rich model zoo
Based on ImageNet1k dataset, PaddleClas provides 23 series of image classification networks such as ResNet, ResNet_vd, Res2Net, HRNet, and MobileNetV3 with brief introductions, reproduction configurations and training tricks. At the same time, the corresponding 117 image classification pretrained models are also available. The GPU inference time of the server-side models are evaluated based on TensorRT. The CPU inference time and storage size of the mobile-side models are evaluated on the Snapdragon 855 (SD855). For more detailed information on the supported pretrained models and their download links, please refer to [**models introduction tutorial**](https://paddleclas.readthedocs.io/zh_CN/latest/models/models_intro.html).
Based on ImageNet1k dataset, PaddleClas provides 23 series of image classification networks such as ResNet, ResNet_vd, Res2Net, HRNet, and MobileNetV3 with brief introductions, reproduction configurations and training tricks. At the same time, the corresponding 117 image classification pretrained models are also available. The GPU inference time of the server-side models are evaluated based on TensorRT. The CPU inference time and storage size of the mobile-side models are evaluated on the Snapdragon 855 (SD855). For more detailed information on the supported pretrained models and their download links, please refer to [**models introduction tutorial**](https://paddleclas-en.readthedocs.io/en/latest/models/models_intro_en.html).
<div align="center">
<img src="./docs/images/models/V100_benchmark/v100.fp32.bs1.main_fps_top1_s.jpg" width="700">
</div>
The above figure shows some of the latest server-side pretrained models. It can be seen from the figure that when using V100 GPU with FP32 and TensorRT, the `Top1` accuracy of the ResNet50_vd_ssld pretrained model on ImageNet1k-val dataset is **82.4%** and that of ResNet101_vd_ssld pretrained model is 83.7%. These pretained models are obtained from SSLD knowledge distillation solution provided by PaddleClas. The marks of the same color and symbol in the figure represent models of different model sizes in the same series. For the introduction of different models, FLOPS, Params and detailed GPU inference time (including the inference speed of T4 GPU with different batch size), please refer to the documentation tutorial for more details: [https://paddleclas.readthedocs.io/zh_CN/latest/models/models_intro.html](https://paddleclas.readthedocs.io/zh_CN/latest/models/models_intro.html)
The above figure shows some of the latest server-side pretrained models. It can be seen from the figure that when using V100 GPU with FP32 and TensorRT, the `Top1` accuracy of the ResNet50_vd_ssld pretrained model on ImageNet1k-val dataset is **83.0%** and that of ResNet101_vd_ssld pretrained model is 83.7%. These pretained models are obtained from SSLD knowledge distillation solution provided by PaddleClas. The marks of the same color and symbol in the figure represent models of different model sizes in the same series. For the introduction of different models, FLOPS, Params and detailed GPU inference time (including the inference speed of T4 GPU with different batch size), please refer to the documentation tutorial for more details: [https://paddleclas-en.readthedocs.io/en/latest/models/models_intro_en.html](https://paddleclas-en.readthedocs.io/en/latest/models/models_intro_en.html)
<div align="center">
......@@ -34,7 +32,7 @@ src="./docs/images/models/mobile_arm_top1.png" width="700">
</div>
The above figure shows the performance of some commonly used mobile-side models, including MobileNetV1, MobileNetV2, MobileNetV3 and ShuffleNetV2 series. The inference time is tested on Snapdragon 855 (SD855) with the batch size set as 1. The `Top1` accuracy of the MV3_large_x1_0_ssld, MV3_small_x1_0_ssld, MV1_ssld and MV2_ssld pretrained model on ImageNet1k-val dataset are 79%, 71.3%, 76.74%, 77.89%, respectively (M is short for MobileNet). MV3_large_x1_0_ssld_int8 is a quantizatied pretrained model for MV3_large_x1_0. More details about the mobile-side models can be seen in [**models introduction tutorial**](https://paddleclas.readthedocs.io/zh_CN/latest/models/models_intro.html)
The above figure shows the performance of some commonly used mobile-side models, including MobileNetV1, MobileNetV2, MobileNetV3 and ShuffleNetV2 series. The inference time is tested on Snapdragon 855 (SD855) with the batch size set as 1. The `Top1` accuracy of the MV3_large_x1_0_ssld, MV3_small_x1_0_ssld, MV1_ssld and MV2_ssld pretrained model on ImageNet1k-val dataset are 79%, 71.3%, 76.74%, 77.89%, respectively (M is short for MobileNet). MV3_large_x1_0_ssld_int8 is a quantizatied pretrained model for MV3_large_x1_0. More details about the mobile-side models can be seen in [**models introduction tutorial**](https://paddleclas-en.readthedocs.io/en/latest/models/models_intro_en.html)
- TODO
- [ ] Reproduction and performance evaluation of EfficientLite, GhostNet, RegNet and ResNeSt.
......@@ -54,7 +52,7 @@ Knowledge distillation refers to using the teacher model to guide the student mo
src="./docs/images/distillation/distillation_perform_s.jpg" width="700">
</div>
Taking the ImageNet1k dataset as an example, the following figure shows the SSLD knowledge distillation method framework. The key points of the method include the choice of teacher model, loss calculation method, iteration number, use of unlabeled data, and ImageNet1k dataset finetune. For detailed introduction and experiments, please refer to [**knowledge distillation tutorial**](https://paddleclas.readthedocs.io/zh_CN/latest/advanced_tutorials/distillation/index.html)
Taking the ImageNet1k dataset as an example, the following figure shows the SSLD knowledge distillation method framework. The key points of the method include the choice of teacher model, loss calculation method, iteration number, use of unlabeled data, and ImageNet1k dataset finetune. For detailed introduction and experiments, please refer to [**knowledge distillation tutorial**](https://paddleclas-en.readthedocs.io/en/latest/advanced_tutorials/distillation/distillation_en.html)
<div align="center">
<img
......@@ -72,7 +70,7 @@ src="./docs/images/image_aug/image_aug_samples_s_en.jpg" width="800">
</div>
PaddleClas provides the reproduction of the above 8 data augmentation algorithms and the evaluation of the effect in a unified environment. The following figure shows the performance of different data augmentation methods based on ResNet50. Compared with the standard transformation, using data augmentation, the recognition accuracy can be increased by up to 1%. For more detailed introduction of data augmentation methods, please refer to the [**data augmentation tutorial**](https://paddleclas.readthedocs.io/zh_CN/latest/advanced_tutorials/image_augmentation/index.html).
PaddleClas provides the reproduction of the above 8 data augmentation algorithms and the evaluation of the effect in a unified environment. The following figure shows the performance of different data augmentation methods based on ResNet50. Compared with the standard transformation, using data augmentation, the recognition accuracy can be increased by up to 1%. For more detailed introduction of data augmentation methods, please refer to the [**data augmentation tutorial**](https://paddleclas-en.readthedocs.io/en/latest/advanced_tutorials/image_augmentation/ImageAugment_en.html).
<div align="center">
......@@ -83,12 +81,12 @@ src="./docs/images/image_aug/main_image_aug_s.jpg" width="600">
## Quick start
Based on flowers102 dataset, one can easily experience different networks, pretrained models and SSLD knowledge distillation method in PaddleClas. More details can be seen in [**Quick start PaddleClas in 30 minutes**](https://paddleclas.readthedocs.io/zh_CN/latest/tutorials/quick_start.html).
Based on flowers102 dataset, one can easily experience different networks, pretrained models and SSLD knowledge distillation method in PaddleClas. More details can be seen in [**Quick start PaddleClas in 30 minutes**](https://paddleclas-en.readthedocs.io/en/latest/tutorials/quick_start_en.html).
## Getting started
For installation, model training, inference, evaluation and finetune in PaddleClas, you can refer to [**gettting started tutorial**](https://paddleclas.readthedocs.io/zh_CN/latest/tutorials/index.html).
For installation, model training, inference, evaluation and finetune in PaddleClas, you can refer to [**gettting started tutorial**](https://paddleclas-en.readthedocs.io/en/latest/tutorials/index.html).
## Featured extension and application
......@@ -107,12 +105,12 @@ The models trained on ImageNet1K dataset are often used as pretrained models for
| Geology | class_num:4<br/>train/val:671/296 | 0.5719 | 0.6781 |
The 100,000 categories' pretrained model can be downloaded here: [download link](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_10w_pretrained.tar). More details can be seen in [**Transfer learning tutorial**](https://paddleclas.readthedocs.io/zh_CN/latest/application/transfer_learning.html).
The 100,000 categories' pretrained model can be downloaded here: [download link](https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_10w_pretrained.tar). More details can be seen in [**Transfer learning tutorial**](https://paddleclas-en.readthedocs.io/en/latest/application/transfer_learning_en.html).
### Object detection
In recent years, object detection tasks attract a lot of attention in academia and industry. The ImageNet classification model is often used for pretrained model in object detection, which can directly affect the effect of object detection. Based on 82.39% ResNet50_vd pretrained model, PaddleDetection provides a Practical Server-side Detection solution, PSS-DET. The solution contains many strategies that can effectively improve the performance while taking limited extra computation cost, such as model pruning, better pretrained model, deformable convolution, cascade rcnn, autoaugment, libra sampling and multi-scale training. Compared with the 79.12% ImageNet1k pretrained model, the 82.39% model can help improve the COCO mAP by 1.5% without any computation cost. Using PSS-DET, the inference speed on single V100 GPU can reach 20FPS when COCO mAP is 47.8%, and reach 61FPS when COCO mAP is 41.6%. For more details, please refer to [**Object Detection tutorial**](https://paddleclas.readthedocs.io/zh_CN/latest/application/object_detection.html).
In recent years, object detection tasks attract a lot of attention in academia and industry. The ImageNet classification model is often used for pretrained model in object detection, which can directly affect the effect of object detection. Based on 82.39% ResNet50_vd pretrained model, PaddleDetection provides a Practical Server-side Detection solution, PSS-DET. The solution contains many strategies that can effectively improve the performance while taking limited extra computation cost, such as model pruning, better pretrained model, deformable convolution, cascade rcnn, autoaugment, libra sampling and multi-scale training. Compared with the 79.12% ImageNet1k pretrained model, the 82.39% model can help improve the COCO mAP by 1.5% without any computation cost. Using PSS-DET, the inference speed on single V100 GPU can reach 20FPS when COCO mAP is 47.8%, and reach 61FPS when COCO mAP is 41.6%. For more details, please refer to [**Object Detection tutorial**](https://paddleclas-en.readthedocs.io/en/latest/application/object_detection_en.html).
- TODO
......
configs/GhostNet/GhostNet_x0_5.yaml 0 → 100644
浏览文件 @ f5168a7a
mode: 'train'
ARCHITECTURE:
name: 'GhostNet_x0_5'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 360
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: 0.1
LEARNING_RATE:
function: 'CosineWarmup'
params:
lr: 0.8
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.0000400
TRAIN:
batch_size: 2048
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
configs/GhostNet/GhostNet_x1_0.yaml 0 → 100644
浏览文件 @ f5168a7a
mode: 'train'
ARCHITECTURE:
name: 'GhostNet_x1_0'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 360
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: 0.1
LEARNING_RATE:
function: 'CosineWarmup'
params:
lr: 0.4
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.0000400
TRAIN:
batch_size: 1024
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
configs/GhostNet/GhostNet_x1_3.yaml 0 → 100644
浏览文件 @ f5168a7a
mode: 'train'
ARCHITECTURE:
name: 'GhostNet_x1_3'
pretrained_model: ""
model_save_dir: "./output/"
classes_num: 1000
total_images: 1281167
save_interval: 1
validate: True
valid_interval: 1
epochs: 360
topk: 5
image_shape: [3, 224, 224]
use_mix: False
ls_epsilon: 0.1
LEARNING_RATE:
function: 'CosineWarmup'
params:
lr: 0.4
OPTIMIZER:
function: 'Momentum'
params:
momentum: 0.9
regularizer:
function: 'L2'
factor: 0.0000400
TRAIN:
batch_size: 1024
num_workers: 4
file_list: "./dataset/ILSVRC2012/train_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- RandCropImage:
size: 224
- RandFlipImage:
flip_code: 1
- AutoAugment:
- NormalizeImage:
scale: 1./255.
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
VALID:
batch_size: 64
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
docs/en/advanced_tutorials/distillation/distillation_en.md
浏览文件 @ f5168a7a
# 1. Introduction of model compression methods
# Introduction of model compression methods
In recent years, deep neural networks have been proven to be an extremely effective method to solve problems in the fields of computer vision and natural language processing. The deep learning methods performs better than traditional methods with suitable network structure and training process.
......@@ -10,12 +10,12 @@ With enough training data, increasing parameters of the neural network by buildi
Parameter redundancy exists in deep neural networks. There are several methods to compress the model suck as pruning ,quantization, knowledge distillation, etc. Knowledge distillation refers to using the teacher model to guide the student model to learn specific tasks, ensuring that the small model has a relatively large effect improvement with the computation cost unchanged, and even obtains similar accuracy with the large model [1]. Combining some of the existing distillation methods [2,3], PaddleClas provides a simple semi-supervised label knowledge distillation solution (SSLD). Top-1 Accuarcy on ImageNet1k dataset has an improvement of more than 3% based on ResNet_vd and MobileNet series, which can be shown as below.
![](../../../images/distillation/distillation_perform.png)
![](../../../images/distillation/distillation_perform_s.jpg)
# 2. SSLD
# SSLD
## 2.1 Itroduction
## Introduction
The following figure shows the framework of SSLD.
......@@ -34,7 +34,7 @@ First, we select nearly 4 million images from ImageNet22k dataset, and integrate
* ImageNet1k finetune. ImageNet1k training set is used for finetuning, which brings a 0.4% accuracy improvement (`75.8%-> 78.9%`).
## 2.2 Data selection
## Data selection
* An important feature of the SSLD distillation scheme is no need for labeled images, so the dataset size can be arbitrarily expanded. Considering the limitation of computing resources, we here only expand the training set of the distillation task based on the ImageNet22k dataset. For SSLD, we used the `Top-k per class` data sampling scheme [3]. Specific steps are as follows.
     * Deduplication of training set. We first deduplicate the ImageNet22k dataset and the ImageNet1k validation set based on the SIFT feature similarity matching method to prevent the added ImageNet22k training set from containing the ImageNet1k validation set images. Finally we removed 4511 similar images. Similar pictures with partial filtering are shown below.
......@@ -45,11 +45,11 @@ First, we select nearly 4 million images from ImageNet22k dataset, and integrate
     * Top-k data selection. There contains 1000 categories in ImageNet1k dataset. For each category, we find out images in the category with Top-k highest score, and finally generate a dataset whose image number does not exceed `1000 * k` (For some categories, there may contain less than k images).
     * The selected images are merged with the ImageNet1k training set to form the new dataset used for the final distillation model training, which contains 5 million images in all.
# 3. Experiments
# Experiments
The distillation solution that PaddleClas provides is combining common training with finetuning. Given a suitable teacher model, the large dataset(5 million) is used for common training and the ImageNet1k dataset is used for finetuning.
## 3.1 Choice of teacher model
## Choice of teacher model
In order to verify the influence of the model size difference between the teacher model and the student model on the distillation results as well as the teacher model accuracy, we conducted several experiments. The training strategy is unified as follows: `cosine_decay_warmup, lr = 1.3, epoch = 120, bs = 2048`, and the student models are all trained from scratch.
......@@ -70,7 +70,7 @@ It can be shown from the table that:
Therefore, during distillation, for the ResNet series student model, we use `ResNeXt101_32x16d_wsl` as the teacher model; for the MobileNet series student model, we use` ResNet50_vd_SSLD` as the teacher model.
## 3.2 Distillation using large-scale dataset
## Distillation using large-scale dataset
Training process is carried out on the large-scale dataset with 5 million images. Specifically, the following table shows more details of different models.
......@@ -83,7 +83,7 @@ Training process is carried out on the large-scale dataset with 5 million images
| ResNet50_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 82.07% |
| ResNet101_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 83.41% |
## 3.3 finetuning using ImageNet1k
## finetuning using ImageNet1k
Finetuning is carried out on ImageNet1k dataset to restore distribution between training set and test set. the following table shows more details of finetuning.
......@@ -97,10 +97,16 @@ Finetuning is carried out on ImageNet1k dataset to restore distribution between
| ResNet50_vd | 60 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 82.39% |
| ResNet101_vd | 30 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 83.73% |
## Data agmentation and Fix strategy
# 4. Application of the distillation model
* Based on experiments mentioned above, we add AutoAugment [4] during training process, and reduced l2_decay from 4e-5 t 2e-5. Finally, the Top-1 accuracy on ImageNet1k dataset can reach 82.99%, with 0.6% improvement compared to the standard SSLD distillation strategy.
## 4.1 Instructions
* For image classsification tasks, The model accuracy can be further improved when the test scale is 1.15 times that of training[5]. For the 82.99% ResNet50_vd pretrained model, it comes to 83.7% using 320x320 for the evaluation. We use Fix strategy to finetune the model with the training scale set as 320x320. During the process, the pre-preocessing pipeline is same for both training and test. All the weights except the fully connected layer are freezed. Finally the top-1 accuracy comes to **84.0%**.
# Application of the distillation model
## Instructions
* Adjust the learning rate of the middle layer. The middle layer feature map of the model obtained by distillation is more refined. Therefore, when the distillation model is used as the pretrained model in other tasks, if the same learning rate as before is adopted, it is easy to destroy the features. If the learning rate of the overall model training is reduced, it will bring about the problem of slow convergence. Therefore, we use the strategy of adjusting the learning rate of the middle layer. specifically:
    * For ResNet50_vd, we set up a learning rate list. The three conv2d convolution parameters before the resiual block have a uniform learning rate multiple, and the four resiual block conv2d have theirs own learning rate parameters, respectively. 5 values need to be set in the list. By the experiment, we find that when used for transfer learning finetune classification model, the learning rate list with `[0.1,0.1,0.2,0.2,0.3]` performs better in most tasks; while in the object detection tasks, `[0.05, 0.05, 0.05, 0.1, 0.15]` can bring greater accuracy gains.
......@@ -108,7 +114,7 @@ Finetuning is carried out on ImageNet1k dataset to restore distribution between
* Appropriate l2 decay. Different l2 decay values are set for different models during training. In order to prevent overfitting, l2 decay is ofen set as large for large models. L2 decay is set as `1e-4` for ResNet50, and `1e-5 ~ 4e-5` for MobileNet series models. L2 decay needs also to be adjusted when applied in other tasks. Taking Faster_RCNN_MobiletNetV3_FPN as an example, we found that only modifying l2 decay can bring up to 0.5% accuracy (mAP) improvement on the COCO2017 dataset.
## 4.2 Transfer learning
## Transfer learning
* To verify the effect of the SSLD pretrained model in transfer learning, we carried out experiments on 10 small datasets. Here, in order to ensure the comparability of the experiment, we use the standard preprocessing process trained by the ImageNet1k dataset. For the distillation model, we also add a simple search method for the learning rate of the middle layers of the distillation pretrained model.
* For ResNet50_vd, the baseline pretrained model Top-1 Acc is 79.12%, the other parameters are got by grid search. For distillation pretrained model, we add learning rate of the middle layers into the search space. The following table shows the results.
......@@ -128,9 +134,8 @@ Finetuning is carried out on ImageNet1k dataset to restore distribution between
* It can be seen that on the above 10 datasets, combined with the appropriate middle layer learning rate, the distillation pretrained model can bring an average accuracy improvement of more than 1%.
## 4.3 Object detection
## Object detection
## 4.3 目标检测
Based on the two-stage Faster/Cascade RCNN model, we verify the effect of the pretrained model obtained by distillation.
......@@ -146,16 +151,16 @@ Training scale and test scale are set as 640x640, and some of the ablationstudie
| Faster RCNN R50_vd FPN | 640/640 | 82.18% | [0.05,0.05,0.1,0.1,0.15] | 36.3% |
It can be seen here that for the baseline pretrained model, excessive adjustment of the middle-layer learning rate actually reduces the performance of the detection model. Based on this distillation model, we also provide a practical server-side detection solution. The detailed configuration and training code are open source, more details can be refer to [PaddleDetection] (https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/rcnn_server_side_det).
It can be seen here that for the baseline pretrained model, excessive adjustment of the middle-layer learning rate actually reduces the performance of the detection model. Based on this distillation model, we also provide a practical server-side detection solution. The detailed configuration and training code are open source, more details can be refer to [PaddleDetection] (https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/rcnn_enhance).
# 五、Practice
# Practice
This section will introduce the SSLD distillation experiments in detail based on the ImageNet-1K dataset. If you want to experience this method quickly, you can refer to [** Quick start PaddleClas in 30 minutes**] (../../tutorials/quick_start.md), whose dataset is set as Flowers102.
## 5.1 Configuration
## Configuration
......@@ -187,7 +192,7 @@ pretrained_model: "./pretrained/ResNet50_vd_ssld_pretrained/"
use_distillation: True
```
## 5.2 Begin to train the network
## Begin to train the network
If everything is ready, users can begin to train the network using the following command.
......@@ -201,7 +206,7 @@ python -m paddle.distributed.launch \
-c ./configs/Distillation/R50_vd_distill_MV3_large_x1_0.yaml
```
## 5.3 Note
## Note
* Before using SSLD, users need to train a teacher model on the target dataset firstly. The teacher model is used to guide the training of the student model.
......@@ -225,3 +230,7 @@ python -m paddle.distributed.launch \
[2] Bagherinezhad H, Horton M, Rastegari M, et al. Label refinery: Improving imagenet classification through label progression[J]. arXiv preprint arXiv:1805.02641, 2018.
[3] Yalniz I Z, Jégou H, Chen K, et al. Billion-scale semi-supervised learning for image classification[J]. arXiv preprint arXiv:1905.00546, 2019.
[4] Cubuk E D, Zoph B, Mane D, et al. Autoaugment: Learning augmentation strategies from data[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2019: 113-123.
[5] Touvron H, Vedaldi A, Douze M, et al. Fixing the train-test resolution discrepancy[C]//Advances in Neural Information Processing Systems. 2019: 8250-8260.
docs/en/advanced_tutorials/distillation/index.rst
浏览文件 @ f5168a7a
......@@ -4,4 +4,4 @@ distillation
.. toctree::
:maxdepth: 3
distillation.md
distillation_en.md
docs/en/advanced_tutorials/image_augmentation/ImageAugment_en.md
浏览文件 @ f5168a7a
# 1. Image Augmentation
# Image Augmentation
Image augmentation is a commonly used regularization method in image classification task, which is often used in scenarios with insufficient data or large model. In this chapter, we mainly introduce 8 image augmentation methods besides standard augmentation methods. Users can apply these methods in their own tasks for better model performance. Under the same conditions, These augmentation methods' performance on ImageNet1k dataset is shown as follows.
......@@ -6,7 +6,7 @@ Image augmentation is a commonly used regularization method in image classificat
![](../../../images/image_aug/main_image_aug.png)
# 2. Common image augmentation methods
# Common image augmentation methods
If without special explanation, all the examples and experiments in this chapter are based on ImageNet1k dataset with the network input image size set as 224.
......@@ -49,11 +49,11 @@ PaddleClas integrates all the above data augmentation strategies. More details i
![](../../../images/image_aug/test_baseline.jpeg)
# 3. Image Transformation
# Image Transformation
Transformation means performing some transformations on the image after `RandCrop`. It mainly contains AutoAugment and RandAugment.
## 3.1 AutoAugment
## AutoAugment
Address：[https://arxiv.org/abs/1805.09501v1](https://arxiv.org/abs/1805.09501v1)
......@@ -89,7 +89,7 @@ The images after `AutoAugment` are as follows.
![][test_autoaugment]
## 3.2 RandAugment
## RandAugment
Address: [https://arxiv.org/pdf/1909.13719.pdf](https://arxiv.org/pdf/1909.13719.pdf)
......@@ -128,7 +128,7 @@ The images after `RandAugment` are as follows.
![][test_randaugment]
# 4. Image Cropping
# Image Cropping
Cropping means performing some transformations on the image after `Transpose`, setting pixels of the cropped area as certain constant. It mainly contains CutOut, RandErasing, HideAndSeek and GridMask.
......@@ -137,7 +137,7 @@ Image cropping methods can be operated before or after normalization. The differ
The above-mentioned cropping transformation ideas are the similar, all to solve the problem of poor generalization ability of the trained model on occlusion images, the difference lies in that their cropping details.
## 4.1 Cutout
## Cutout
Address: [https://arxiv.org/abs/1708.04552](https://arxiv.org/abs/1708.04552)
......@@ -173,7 +173,7 @@ The images after `Cutout` are as follows.
![][test_cutout]
## 4.2 RandomErasing
## RandomErasing
Address: [https://arxiv.org/pdf/1708.04896.pdf](https://arxiv.org/pdf/1708.04896.pdf)
......@@ -211,7 +211,7 @@ The images after `RandomErasing` are as follows.
![][test_randomerassing]
## 4.3 HideAndSeek
## HideAndSeek
Address: [https://arxiv.org/pdf/1811.02545.pdf](https://arxiv.org/pdf/1811.02545.pdf)
......@@ -252,7 +252,7 @@ The images after `HideAndSeek` are as follows.
![][test_hideandseek]
## 4.4 GridMask
## GridMask
Address：[https://arxiv.org/abs/2001.04086](https://arxiv.org/abs/2001.04086)
Github repo：[https://github.com/akuxcw/GridMask](https://github.com/akuxcw/GridMask)
......@@ -307,13 +307,13 @@ The images after `GridMask` are as follows.
![][test_gridmask]
# 5. Image aliasing
# Image aliasing
Aliasing means performing some transformations on the image after `Batch`, which contains Mixup and Cutmix.
Data augmentation methods introduced before are based on single image while aliasing is carried on a certain batch to generate a new batch.
## 5.1 Mixup
## Mixup
Address: [https://arxiv.org/pdf/1710.09412.pdf](https://arxiv.org/pdf/1710.09412.pdf)
......@@ -358,7 +358,7 @@ The images after `Mixup` are as follows.
![][test_mixup]
## 5.2 Cutmix
## Cutmix
Address: [https://arxiv.org/pdf/1905.04899v2.pdf](https://arxiv.org/pdf/1905.04899v2.pdf)
......@@ -402,7 +402,7 @@ The images after `Cutmix` are as follows.
![][test_cutmix]
# 6. Experiments
# Experiments
Based on PaddleClas, Metrics of different augmentation methods on ImageNet1k dataset are as follows.
......@@ -426,15 +426,15 @@ Based on PaddleClas, Metrics of different augmentation methods on ImageNet1k dat
## 7. Data augmentation practice
## Data augmentation practice
Experiments about data augmentation will be introduced in detail in this section. If you want to quickly experience these methods, please refer to [**Quick start PaddleClas in 30 miniutes**](../../tutorials/quick_start_en.md).
## 7.1 Configurations
## Configurations
Since hyperparameters differ from different augmentation methods. For better understanding, we list 8 augmentation configuration files in `configs/DataAugment` based on ResNet50. Users can train the model with `tools/run.sh`. The following are 3 of them.
### 7.1.1 RandAugment
### RandAugment
Configuration of `RandAugment` is shown as follows. `Num_layers`(default as 2) and `magnitude`(default as 5) are two hyperparameters.
......@@ -460,7 +460,7 @@ Configuration of `RandAugment` is shown as follows. `Num_layers`(default as 2) a
- ToCHWImage:
```
### 7.1.2 Cutout
### Cutout
Configuration of `Cutout` is shown as follows. `n_holes`(default as 1) and `n_holes`(default as 112) are two hyperparameters.
......@@ -485,7 +485,7 @@ Configuration of `Cutout` is shown as follows. `n_holes`(default as 1) and `n_ho
- ToCHWImage:
```
### 7.1.3 Mixup
### Mixup
Configuration of `Mixup` is shown as follows. `alpha`(default as 0.2) is hyperparameter which users need to care about. What's more, `use_mix` need to be set as `True` in the root of the configuration.
......@@ -511,7 +511,7 @@ Configuration of `Mixup` is shown as follows. `alpha`(default as 0.2) is hyperpa
alpha: 0.2
```
## 7.2 启动命令
## 启动命令
Users can use the following command to start the training process, which can also be referred to `tools/run.sh`.
......@@ -524,7 +524,7 @@ python -m paddle.distributed.launch \
-c ./configs/DataAugment/ResNet50_Cutout.yaml
```
## 7.3 Note
## Note
* When using augmentation methods based on image aliasing, users need to set `use_mix` in the configuration file as `True`. In addition, because the label needs to be aliased when the image is aliased, the accuracy of the training data cannot be calculated. The training accuracy rate was not printed during the training process.
......
docs/en/advanced_tutorials/image_augmentation/index.rst
浏览文件 @ f5168a7a
......@@ -4,4 +4,4 @@ image_augmentation
.. toctree::
:maxdepth: 3
ImageAugment.md
ImageAugment_en.md
docs/en/application/index.rst
浏览文件 @ f5168a7a
......@@ -4,5 +4,5 @@ application
.. toctree::
:maxdepth: 2
transfer_learning.md
object_detection.md
transfer_learning_en.md
object_detection_en.md
docs/en/application/object_detection_en.md 0 → 100644
浏览文件 @ f5168a7a
# General object detection
## Practical Server-side detection method base on RCNN
### Introduction
* In recent years, object detection tasks have attracted widespread attention. [PaddleClas](https://github.com/PaddlePaddle/PaddleClas) open-sourced the ResNet50_vd_SSLD pretrained model based on ImageNet(Top1 Acc 82.4%). And based on the pretrained model, PaddleDetection provided the PSS-DET (Practical Server-side detection) with the help of the rich operators in PaddleDetection. The inference speed can reach 61FPS on single V100 GPU when COCO mAP is 41.6%, and 20FPS when COCO mAP is 47.8%.
* We take the standard `Faster RCNN ResNet50_vd FPN` as an example. The following table shows ablation study of PSS-DET.
| Trick | Train scale | Test scale | COCO mAP | Infer speed/FPS |
|- |:-: |:-: | :-: | :-: |
| `baseline` | 640x640 | 640x640 | 36.4% | 43.589 |
| +`test proposal=pre/post topk 500/300` | 640x640 | 640x640 | 36.2% | 52.512 |
| +`fpn channel=64` | 640x640 | 640x640 | 35.1% | 67.450 |
| +`ssld pretrain` | 640x640 | 640x640 | 36.3% | 67.450 |
| +`ciou loss` | 640x640 | 640x640 | 37.1% | 67.450 |
| +`DCNv2` | 640x640 | 640x640 | 39.4% | 60.345 |
| +`3x, multi-scale training` | 640x640 | 640x640 | 41.0% | 60.345 |
| +`auto augment` | 640x640 | 640x640 | 41.4% | 60.345 |
| +`libra sampling` | 640x640 | 640x640 | 41.6% | 60.345 |
Based on the ablation experiments, Cascade RCNN and larger inference scale(1000x1500) are used for better performance. The final COCO mAP is 47.8%
and the following figure shows `mAP-Speed` curves for some common detectors.
![pssdet](../../images/det/pssdet.png)
**Note**
> For fair comparison, inference time for PSS-DET models on V100 GPU is transformed to Titan V GPU by multiplying by 1.2 times.
For more detailed information, you can refer to [PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/rcnn_server_side_det).
## Practical Mobile-side detection method base on RCNN
* This part is comming soon!
docs/en/application/transfer_learning_en.md 0 → 100644
浏览文件 @ f5168a7a
# Transfer learning in image classification
Transfer learning is an important part of machine learning, which is widely used in various fields such as text and images. Here we mainly introduce transfer learning in the field of image classification, which is often called domain transfer, such as migration of the ImageNet classification model to the specified image classification task, such as flower classification.
## Hyperparameter search
ImageNet is the widely used dataset for image classification. A series of empirical hyperparameters have been summarized. High accuracy can be got using the hyperparameters. However, when applied in the specified dataset, the hyperparameters may not be optimal. There are two commonly used hyperparameter search methods that can be used to help us obtain better model hyperparameters.
### Grid search
For grid search, which is also called exhaustive search, the optimal value is determined by finding the best solution from all solutions in the search space. The method is simple and effective, but when the search space is large, it takes huge computing resource.
### Bayesian search
Bayesian search, which is also called Bayesian optimization, is realized by randomly selecting a group of hyperparameters in the search space. Gaussian process is used to update the hyperparameters, compute their expected mean and variance according to the performance of the previous hyperparameters. The larger the expected mean, the greater the probability of being close to the optimal solution. The larger the expected variance, the greater the uncertainty. Usually, the hyperparameter point with large expected mean is called `exporitation`, and the hyperparameter point with large variance is called `exploration`. Acquisition function is defined to balance the expected mean and variance. The currently selected hyperparameter point is viewed as the optimal position with maximum probability.
According to the above two search schemes, we carry out some experiments based on fixed scheme and two search schemes on 8 open source datasets. As the experimental scheme in [1], we search for 4 hyperparameters, the search space and The experimental results are as follows:
a fixed set of parameter experiments and two search schemes on 8 open source data sets. With reference to the experimental scheme of [1], we search for 4 hyperparameters, the search space and the experimental results are as follows:
- Fixed scheme.
```
lr=0.003，l2 decay=1e-4，label smoothing=False，mixup=False
```
- Search space of the hyperparameters.
```
lr: [0.1, 0.03, 0.01, 0.003, 0.001, 0.0003, 0.0001]
l2 decay: [1e-3, 3e-4, 1e-4, 3e-5, 1e-5, 3e-6, 1e-6]
label smoothing: [False, True]
mixup: [False, True]
```
It takes 196 times for grid search, and takes 10 times less for Bayesian search. The baseline is trained by using ImageNet1k pretrained model based on ResNet50_vd and fixed scheme. The follow shows the experiments.
| Dataset | Fix scheme | Grid search | Grid search time | Bayesian search | Bayesian search time|
| ------------------ | -------- | -------- | -------- | -------- | ---------- |
| Oxford-IIIT-Pets | 93.64% | 94.55% | 196 | 94.04% | 20 |
| Oxford-102-Flowers | 96.08% | 97.69% | 196 | 97.49% | 20 |
| Food101 | 87.07% | 87.52% | 196 | 87.33% | 23 |
| SUN397 | 63.27% | 64.84% | 196 | 64.55% | 20 |
| Caltech101 | 91.71% | 92.54% | 196 | 92.16% | 14 |
| DTD | 76.87% | 77.53% | 196 | 77.47% | 13 |
| Stanford Cars | 85.14% | 92.72% | 196 | 92.72% | 25 |
| FGVC Aircraft | 80.32% | 88.45% | 196 | 88.36% | 20 |
- The above experiments verify that Bayesian search only reduces the accuracy by 0% to 0.4% under the condition of reducing the number of searches by about 10 times compared to grid search.
- The search space can be expaned easily using Bayesian search.
## Large-scale image classification
In practical applications, due to the lack of training data, the classification model trained on the ImageNet1k data set is often used as the pretrained model for other image classification tasks. In order to further help solve practical problems, based on ResNet50_vd, Baidu open sourced a self-developed large-scale classification pretrained model, in which the training data contains 100,000 categories and 43 million pictures.
We conducted transfer learning experiments on 6 self-collected datasets,
using a set of fixed parameters and a grid search method, in which the number of training rounds was set to 20epochs, the ResNet50_vd model was selected, and the ImageNet pre-training accuracy was 79.12%. The comparison results of the experimental data set parameters and model accuracy are as follows:
Fixed scheme：
```
lr=0.001，l2 decay=1e-4，label smoothing=False，mixup=False
```
| Dataset | Statstics | **Pretrained moel on ImageNet <br />Top-1(fixed)/Top-1(search)** | **Pretrained moel on large-scale dataset<br />Top-1(fixed)/Top-1(search)** |
| --------------- | ----------------------------------------- | -------------------------------------------------------- | --------------------------------------------------------- |
| Flowers | class:102<br />train:5789<br />valid:2396 | 0.7779/0.9883 | 0.9892/0.9954 |
| Hand-painted stick figures | Class:18<br />train:1007<br />valid:432 | 0.8795/0.9196 | 0.9107/0.9219 |
| Leaves | class:6<br />train:5256<br />valid:2278 | 0.8212/0.8482 | 0.8385/0.8659 |
| Container vehicle | Class:115<br />train:4879<br />valid:2094 | 0.6230/0.9556 | 0.9524/0.9702 |
| Chair | class:5<br />train:169<br />valid:78 | 0.8557/0.9688 | 0.9077/0.9792 |
| Geology | class:4<br />train:671<br />valid:296 | 0.5719/0.8094 | 0.6781/0.8219 |
- The above experiments verified that for fixed parameters, compared with the pretrained model on ImageNet, using the large-scale classification model as a pretrained model can help us improve the model performance on a new dataset in most cases. Parameter search can be further helpful to the model performance.
## Reference
[1] Kornblith, Simon, Jonathon Shlens, and Quoc V. Le. "Do better imagenet models transfer better?." *Proceedings of the IEEE conference on computer vision and pattern recognition*. 2019.
[2] Kolesnikov, Alexander, et al. "Large Scale Learning of General Visual Representations for Transfer." *arXiv preprint arXiv:1912.11370* (2019).
docs/en/change_log.md 已删除 100644 → 0
浏览文件 @ f03fab0e
# Release Notes
* 2020.04.14: first commit
docs/en/competition_support_en.md 0 → 100644
浏览文件 @ f5168a7a
### Competition Support
PaddleClas stems from the Baidu's visual business applications and the exploration of frontier visual capabilities. It has helped us achieve leading results in many key events, and continues to promote more frontier visual solutions and landing applications.
* 1st place in 2018 Kaggle Open Images V4 object detection challenge
* 2nd place in 2019 Kaggle Open Images V5 object detection challenge
* The report is avaiable here: [https://arxiv.org/pdf/1911.07171.pdf](https://arxiv.org/pdf/1911.07171.pdf)
* The pretrained model and code is avaiable here: [source code](https://github.com/PaddlePaddle/PaddleDetection/blob/master/docs/featured_model/OIDV5_BASELINE_MODEL.md)
* 2nd place in Kacggle Landmark Retrieval Challenge 2019
* The report is avaiable here: [https://arxiv.org/abs/1906.03990](https://arxiv.org/abs/1906.03990)
* The pretrained model and code is avaiable here: [source code](https://github.com/PaddlePaddle/Research/tree/master/CV/landmark)
* 2nd place in Kaggle Landmark Recognition Challenge 2019
* The report is avaiable here: [https://arxiv.org/abs/1906.03990](https://arxiv.org/abs/1906.03990)
* The pretrained model and code is avaiable here: [source code](https://github.com/PaddlePaddle/Research/tree/master/CV/landmark)
* A-level certificate of three tasks: printed text OCR, face recognition and landmark recognition in the first multimedia information recognition technology competition
docs/en/extension/index.rst
浏览文件 @ f5168a7a
......@@ -4,9 +4,9 @@ extension
.. toctree::
:maxdepth: 1
paddle_inference.md
paddle_mobile_inference.md
paddle_quantization.md
multi_machine_training.md
paddle_hub.md
paddle_serving.md
paddle_inference_en.md
paddle_mobile_inference_en.md
paddle_quantization_en.md
multi_machine_training_en.md
paddle_hub_en.md
paddle_serving_en.md
docs/en/extension/paddle_inference_en.md
浏览文件 @ f5168a7a
# Classification Framework
# Prediction Framework
## I. Introduction
## Introduction
Models for Paddle are stored in many different forms, which can be roughly divided into two categories：
1. persistable model（the models saved by fluid.save_persistables）
......@@ -54,7 +54,7 @@ Regardless of the inference method, it basically includes the following main ste
There are two main differences in different inference methods: building the engine and executing the forecast. The following sections will be introduced in detail
## II. Model Transformation
## Model Transformation
During training, we usually save some checkpoints (persistable models). These are just model weight files and cannot be directly loaded by the prediction engine to predict, so we usually find suitable checkpoints after the training and convert them to inference model. There are two main steps: 1. Build a training engine, 2. Save the inference model, as shown below.
......@@ -95,7 +95,7 @@ python tools/export_model.py \
--o=the saved path of model and params
```
## III. prediction engine + inference model
## Prediction engine + inference model
The complete example is provided in the `tools/infer/predict.py`，just execute the following command to complete the prediction:
......@@ -161,7 +161,7 @@ More parameters information can be refered in [Paddle Python prediction API](htt
By default, Paddle's wheel package does not include the TensorRT prediction engine. If you need to use TensorRT for prediction optimization, you need to compile the corresponding wheel package yourself. For the compilation method, please refer to Paddle's compilation guide. [Paddle compilation](https://www.paddlepaddle.org.cn/documentation/docs/zh/install/compile/fromsource.html)。
## IV、Training engine + persistable model prediction
## Training engine + persistable model prediction
A complete example is provided in the `tools/infer/infer.py`, just execute the following command to complete the prediction：
......@@ -212,7 +212,7 @@ outputs = exe.run(infer_prog,
For the above parameter descriptions, please refer to the official website [fluid.Executor](https://www.paddlepaddle.org.cn/documentation/docs/zh/api_cn/executor_cn/Executor_cn.html)
## V、training engine + inference model prediction
## Training engine + inference model prediction
A complete example is provided in `tools/infer/py_infer.py`, just execute the following command to complete the prediction：
......
docs/en/extension/paddle_mobile_inference_en.md
浏览文件 @ f5168a7a
# Paddle-Lite
## I. Introduction
## Introduction
[Paddle-Lite](https://github.com/PaddlePaddle/Paddle-Lite) is a set of lightweight inference engine which is fully functional, easy to use and then performs well. Lightweighting is reflected in the use of fewer bits to represent the weight and activation of the neural network, which can greatly reduce the size of the model, solve the problem of limited storage space of the mobile device, and the inference speed is better than other frameworks on the whole.
In [PaddleClas](https://github.com/PaddlePaddle/PaddleClas), we uses Paddle-Lite to [evaluate the performance on the mobile device](../models/Mobile.md), in this section we uses the `MobileNetV1` model trained on the `ImageNet1k` dataset as an example to introduce how to use `Paddle-Lite` to evaluate the model speed on the mobile terminal (evaluated on SD855)
## II. Evaluation Steps
## Evaluation Steps
### I. Export the Inference Model
### Export the Inference Model
* First you should transform the saved model during training to the special model which can be used to inference, the special model can be exported by `tools/export_model.py`, the specific way of transform is as follows.
......@@ -19,7 +19,7 @@ python tools/export_model.py -m MobileNetV1 -p pretrained/MobileNetV1_pretrained
Finally the `model` and `parmas` can be saved in `inference/MobileNetV1`.
### II. Download Benchmark Binary File
### Download Benchmark Binary File
* Use the adb (Android Debug Bridge) tool to connect the Android phone and the PC, then develop and debug. After installing adb and ensuring that the PC and the phone are successfully connected, use the following command to view the ARM version of the phone and select the pre-compiled library based on ARM version.
......@@ -39,7 +39,7 @@ If the ARM version is v7, the v7 benchmark_bin file should be downloaded, the co
wget -c https://paddle-inference-dist.bj.bcebos.com/PaddleLite/benchmark_0/benchmark_bin_v7
```
### III. the Inference Speeds
### Inference benchmark
After the PC and mobile phone are successfully connected, use the following command to start the model evaluation.
......@@ -63,7 +63,7 @@ MobileNetV1 min = 10.03200 max = 9.94300 averag
Here is the model inference speed under different number of threads, the unit is FPS, taking model on one threads as an example, the average speed of MobileNetV1 on SD855 is `30.79750FPS`.
### IV. Model Optimization and Speed Evaluation
### Model Optimization and Speed Evaluation
* In II.III section, we mention that the model will be optimized before evaluation, here you can first optimize the model, and then directly load the optimized model for speed evaluation
......
docs/en/extension/paddle_quantization_en.md 0 → 100644
浏览文件 @ f5168a7a
# Model Quantifization
Int8 quantization is one of the key features in [PaddleSlim](https://github.com/PaddlePaddle/PaddleSlim).
It supports two kinds of training aware, **Dynamic strategy** and **Static strategy**,
layer-wise and channel-wise quantization,
and using PaddleLite to deploy models generated by PaddleSlim.
By using this toolkit, [PaddleClas](https://github.com/PaddlePaddle/PaddleClas) quantized the mobilenet_v3_large_x1_0 model whose accuracy is 78.9% after distilled.
After quantized, the prediction speed is accelerated from 19.308ms to 14.395ms on SD855.
The storage size is reduced from 21M to 10M.
The top1 recognition accuracy rate is 75.9%.
For specific training methods, please refer to [PaddleSlim quant aware](https://paddlepaddle.github.io/PaddleSlim/quick_start/quant_aware_tutorial.html)。
docs/en/extension/paddle_serving_en.md 0 → 100644
浏览文件 @ f5168a7a
# Model Service Deployment
## Overview
[Paddle Serving](https://github.com/PaddlePaddle/Serving) aims to help deep-learning researchers to easily deploy online inference services, supporting one-click deployment of industry, high concurrency and efficient communication between client and server and supporting multiple programming languages to develop clients.
Taking HTTP inference service deployment as an example to introduce how to use PaddleServing to deploy model services in PaddleClas.
## Serving Install
It is recommends to use docker to install and deploy the Serving environment in the Serving official website, first, you need to pull the docker environment and create Serving-based docker.
```shell
nvidia-docker pull hub.baidubce.com/paddlepaddle/serving:0.2.0-gpu
nvidia-docker run -p 9292:9292 --name test -dit hub.baidubce.com/paddlepaddle/serving:0.2.0-gpu
nvidia-docker exec -it test bash
```
In docker, you need to install some packages about Serving
```shell
pip install paddlepaddle-gpu
pip install paddle-serving-client
pip install paddle-serving-server-gpu
```
* If the installation speed is too slow, you can add `-i https://pypi.tuna.tsinghua.edu.cn/simple` following pip to speed up the process.
* If you want to deploy CPU service, you can install the cpu version of Serving, the command is as follow.
```shell
pip install paddle-serving-server
```
### Export Model
Exporting the Serving model using `tools/export_serving_model.py`, taking ResNet50_vd as an example, the command is as follow.
```shell
python tools/export_serving_model.py -m ResNet50_vd -p ./pretrained/ResNet50_vd_pretrained/ -o serving
```
finally, the client configures, model parameters and structure file will be saved in `ppcls_client_conf` and `ppcls_model`.
### Service Deployment and Request
* Using the following commands to start the Serving.
```shell
python tools/serving/image_service_gpu.py serving/ppcls_model workdir 9292
```
`serving/ppcls_model` is the address of the Serving model just saved, `workdir` is the work directory, and `9292` is the port of the service.
* Using the following script to send an identification request to the Serving and return the result.
```
python tools/serving/image_http_client.py 9292 ./docs/images/logo.png
```
`9292` is the port for sending the request, which is consistent with the Serving starting port, and `./docs/images/logo.png` is the test image, the final top1 label and probability are returned.
* For more Serving deployment, such RPC inference service, you can refer to the Serving official website: [https://github.com/PaddlePaddle/Serving/tree/develop/python/examples/imagenet](https://github.com/PaddlePaddle/Serving/tree/develop/python/examples/imagenet)
docs/en/faq_en.md 0 → 100644
浏览文件 @ f5168a7a
# FAQ
>>
* Why are the metrics different for different cards?
* A: Fleet is the default option for the use of PaddleClas. Each GPU card is taken as a single trainer and deals with different images, which cause the final small difference. Single card evalution is suggested to get the accurate results if you use `tools/eval.py`. You can also use `tools/eval_multi_platform.py` to evalute the models on multiple GPU cards, which is also supported on Windows and CPU.
>>
* Q: Why `Mixup` or `Cutmix` is not used even if I have already add the data operation in the configuration file?
* A: When using `Mixup` or `Cutmix`, you also need to add `use_mix: True` in the configuration file to make it work properly.
>>
* Q: During evaluation and inference, pretrained model address is assgined, but the weights can not be imported. Why?
* A: Prefix of the pretrained model is needed. For example, if the pretained weights are located in `output/ResNet50_vd/19`, with the filename `output/ResNet50_vd/19/ppcls.pdparams`, then `pretrained_model` in the configuration file needs to be `output/ResNet50_vd/19/ppcls`.
>>
* Q: Why are the metrics 0.3% lower than that shown in the model zoo for `EfficientNet` series of models?
* A: Resize method is set as `Cubic` for `EfficientNet`(interpolation is set as 2 in OpenCV), while other models are set as `Bilinear`(interpolation is set as None in OpenCV). Therefore, you need to modify the interpolation explicitly in `ResizeImage`. Specifically, the following configuration is a demo for EfficientNet.
```
VALID:
batch_size: 16
num_workers: 4
file_list: "./dataset/ILSVRC2012/val_list.txt"
data_dir: "./dataset/ILSVRC2012/"
shuffle_seed: 0
transforms:
- DecodeImage:
to_rgb: True
to_np: False
channel_first: False
- ResizeImage:
resize_short: 256
interpolation: 2
- CropImage:
size: 224
- NormalizeImage:
scale: 1.0/255.0
mean: [0.485, 0.456, 0.406]
std: [0.229, 0.224, 0.225]
order: ''
- ToCHWImage:
```
>>
* Q: What should I do if I want to transform the weights' format from `pdparams` to an earlier version(before Paddle1.7.0), which consists of the scattered files?
* A: You can use `fluid.load` to load the `pdparams` weights and use `fluid.io.save_vars` to save the weights as scattered files.
docs/en/index.rst
浏览文件 @ f5168a7a
......@@ -11,9 +11,7 @@ Welcome to PaddleClas！
advanced_tutorials/index
application/index
extension/index
competition_support.md
model_zoo.md
change_log.md
faq.md
competition_support_en.md
update_history_en.md
faq_en.md
:math:`PaddlePaddle2020`
docs/en/models/DPN_DenseNet_en.md 0 → 100644
浏览文件 @ f5168a7a
# DPN and DenseNet series
## Overview
DenseNet is a new network structure proposed in 2017 and was the best paper of CVPR. The network has designed a new cross-layer connected block called dense-block. Compared to the bottleneck in ResNet, dense-block has designed a more aggressive dense connection module, that is, connecting all the layers to each other, and each layer will accept all the layers in front of it as its additional input. DenseNet stacks all dense-blocks into a densely connected network. The dense connection makes DenseNet easier to backpropagate, making the network easier to train and converge. The full name of DPN is Dual Path Networks, which is a network composed of DenseNet and ResNeXt, which proves that DenseNet can extract new features from the previous level, and ResNeXt essentially reuses the extracted features . The author further analyzes and finds that ResNeXt has high reuse rate for features, but low redundancy, while DenseNet can create new features, but with high redundancy. Combining the advantages of the two structures, the author designed the DPN network. In the end, the DPN network achieved better results than ResNeXt and DenseNet under the same FLOPS and parameters.
The FLOPS, parameters, and inference time on the T4 GPU of this series of models are shown in the figure below.
![](../../images/models/T4_benchmark/t4.fp32.bs4.DPN.flops.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.DPN.params.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.DPN.png)
![](../../images/models/T4_benchmark/t4.fp16.bs4.DPN.png)
The pretrained models of these two types of models (a total of 10) are open sourced in PaddleClas at present. The indicators are shown in the figure above. It is easy to observe that under the same FLOPS and parameters, DPN has higher accuracy than DenseNet. However,because DPN has more branches, its inference speed is slower than DenseNet. Since DenseNet264 has the deepest layers in all DenseNet networks, it has the largest parameters,DenseNet161 has the largest width, resulting the largest FLOPs and the highest accuracy in this series. From the perspective of inference speed, DenseNet161, which has a large FLOPs and high accuracy, has a faster speed than DenseNet264, so it has a greater advantage than DenseNet264.
For DPN series networks, the larger the model's FLOPs and parameters, the higher the model's accuracy. Among them, since the width of DPN107 is the largest, it has the largest number of parameters and FLOPs in this series of networks.
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Parameters<br>(M) |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| DenseNet121 | 0.757 | 0.926 | 0.750 | | 5.690 | 7.980 |
| DenseNet161 | 0.786 | 0.941 | 0.778 | | 15.490 | 28.680 |
| DenseNet169 | 0.768 | 0.933 | 0.764 | | 6.740 | 14.150 |
| DenseNet201 | 0.776 | 0.937 | 0.775 | | 8.610 | 20.010 |
| DenseNet264 | 0.780 | 0.939 | 0.779 | | 11.540 | 33.370 |
| DPN68 | 0.768 | 0.934 | 0.764 | 0.931 | 4.030 | 10.780 |
| DPN92 | 0.799 | 0.948 | 0.793 | 0.946 | 12.540 | 36.290 |
| DPN98 | 0.806 | 0.951 | 0.799 | 0.949 | 22.220 | 58.460 |
| DPN107 | 0.809 | 0.953 | 0.802 | 0.951 | 35.060 | 82.970 |
| DPN131 | 0.807 | 0.951 | 0.801 | 0.949 | 30.510 | 75.360 |
## Inference speed based on V100 GPU
| Models | Crop Size | Resize Short Size | FP32<br>Batch Size=1<br>(ms) |
|-------------|-----------|-------------------|--------------------------|
| DenseNet121 | 224 | 256 | 4.371 |
| DenseNet161 | 224 | 256 | 8.863 |
| DenseNet169 | 224 | 256 | 6.391 |
| DenseNet201 | 224 | 256 | 8.173 |
| DenseNet264 | 224 | 256 | 11.942 |
| DPN68 | 224 | 256 | 11.805 |
| DPN92 | 224 | 256 | 17.840 |
| DPN98 | 224 | 256 | 21.057 |
| DPN107 | 224 | 256 | 28.685 |
| DPN131 | 224 | 256 | 28.083 |
## Inference speed based on T4 GPU
| Models | Crop Size | Resize Short Size | FP16<br>Batch Size=1<br>(ms) | FP16<br>Batch Size=4<br>(ms) | FP16<br>Batch Size=8<br>(ms) | FP32<br>Batch Size=1<br>(ms) | FP32<br>Batch Size=4<br>(ms) | FP32<br>Batch Size=8<br>(ms) |
|-------------|-----------|-------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
| DenseNet121 | 224 | 256 | 4.16436 | 7.2126 | 10.50221 | 4.40447 | 9.32623 | 15.25175 |
| DenseNet161 | 224 | 256 | 9.27249 | 14.25326 | 20.19849 | 10.39152 | 22.15555 | 35.78443 |
| DenseNet169 | 224 | 256 | 6.11395 | 10.28747 | 13.68717 | 6.43598 | 12.98832 | 20.41964 |
| DenseNet201 | 224 | 256 | 7.9617 | 13.4171 | 17.41949 | 8.20652 | 17.45838 | 27.06309 |
| DenseNet264 | 224 | 256 | 11.70074 | 19.69375 | 24.79545 | 12.14722 | 26.27707 | 40.01905 |
| DPN68 | 224 | 256 | 11.7827 | 13.12652 | 16.19213 | 11.64915 | 12.82807 | 18.57113 |
| DPN92 | 224 | 256 | 18.56026 | 20.35983 | 29.89544 | 18.15746 | 23.87545 | 38.68821 |
| DPN98 | 224 | 256 | 21.70508 | 24.7755 | 40.93595 | 21.18196 | 33.23925 | 62.77751 |
| DPN107 | 224 | 256 | 27.84462 | 34.83217 | 60.67903 | 27.62046 | 52.65353 | 100.11721 |
| DPN131 | 224 | 256 | 28.58941 | 33.01078 | 55.65146 | 28.33119 | 46.19439 | 89.24904 |
docs/en/models/EfficientNet_and_ResNeXt101_wsl_en.md 0 → 100644
浏览文件 @ f5168a7a
# EfficientNet and ResNeXt101_wsl series
## Overview
EfficientNet is a lightweight NAS-based network released by Google in 2019. EfficientNetB7 refreshed the classification accuracy of ImageNet-1k at that time. In this paper, the author points out that the traditional methods to improve the performance of neural networks mainly start with the width of the network, the depth of the network, and the resolution of the input picture.
However, the author found that balancing these three dimensions is essential for improving accuracy and efficiency through experiments.
Therefore, the author summarized how to balance the three dimensions at the same time through a series of experiments.
At the same time, based on this scaling method, the author built a total of 7 networks B1-B7 in the EfficientNet series on the basis of EfficientNetB0, and with the same FLOPS and parameters, the accuracy reached state-of-the-art effect.
ResNeXt is an improved version of ResNet that proposed by Facebook in 2016. In 2019, Facebook researchers studied the accuracy limit of the series network on ImageNet through weakly-supervised-learning. In order to distinguish the previous ResNeXt network, the suffix of this series network is WSL, where WSL is the abbreviation of weakly-supervised-learning. In order to have stronger feature extraction capability, the researchers further enlarged the network width, among which the largest ResNeXt101_32x48d_wsl has 800 million parameters. It was trained under 940 million weak-labeled images, and the results were finetune trained on imagenet-1k. Finally, the acc-1 of imagenet-1k reaches 85.4%, which is also the network with the highest precision under the resolution of 224x224 on imagenet-1k so far. In Fix-ResNeXt, the author used a larger image resolution, made a special Fix strategy for the inconsistency of image data preprocessing in training and testing, and made ResNeXt101_32x48d_wsl have a higher accuracy. Since it used the Fix strategy, it was named Fix-ResNeXt101_32x48d_wsl.
The FLOPS, parameters, and inference time on the T4 GPU of this series of models are shown in the figure below.
![](../../images/models/T4_benchmark/t4.fp32.bs4.EfficientNet.flops.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.EfficientNet.params.png)
![](../../images/models/T4_benchmark/t4.fp32.bs1.EfficientNet.png)
![](../../images/models/T4_benchmark/t4.fp16.bs1.EfficientNet.png)
At present, there are a total of 14 pretrained models of the two types of models that PaddleClas open source. It can be seen from the above figure that the advantages of the EfficientNet series network are very obvious. The ResNeXt101_wsl series model uses more data, and the final accuracy is also higher. EfficientNet_B0_small removes SE_block based on EfficientNet_B0, which has faster inference speed.
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Parameters<br>(M) |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| ResNeXt101_<br>32x8d_wsl | 0.826 | 0.967 | 0.822 | 0.964 | 29.140 | 78.440 |
| ResNeXt101_<br>32x16d_wsl | 0.842 | 0.973 | 0.842 | 0.972 | 57.550 | 152.660 |
| ResNeXt101_<br>32x32d_wsl | 0.850 | 0.976 | 0.851 | 0.975 | 115.170 | 303.110 |
| ResNeXt101_<br>32x48d_wsl | 0.854 | 0.977 | 0.854 | 0.976 | 173.580 | 456.200 |
| Fix_ResNeXt101_<br>32x48d_wsl | 0.863 | 0.980 | 0.864 | 0.980 | 354.230 | 456.200 |
| EfficientNetB0 | 0.774 | 0.933 | 0.773 | 0.935 | 0.720 | 5.100 |
| EfficientNetB1 | 0.792 | 0.944 | 0.792 | 0.945 | 1.270 | 7.520 |
| EfficientNetB2 | 0.799 | 0.947 | 0.803 | 0.950 | 1.850 | 8.810 |
| EfficientNetB3 | 0.812 | 0.954 | 0.817 | 0.956 | 3.430 | 11.840 |
| EfficientNetB4 | 0.829 | 0.962 | 0.830 | 0.963 | 8.290 | 18.760 |
| EfficientNetB5 | 0.836 | 0.967 | 0.837 | 0.967 | 19.510 | 29.610 |
| EfficientNetB6 | 0.840 | 0.969 | 0.842 | 0.968 | 36.270 | 42.000 |
| EfficientNetB7 | 0.843 | 0.969 | 0.844 | 0.971 | 72.350 | 64.920 |
| EfficientNetB0_<br>small | 0.758 | 0.926 | | | 0.720 | 4.650 |
## Inference speed based on V100 GPU
| Models | Crop Size | Resize Short Size | FP32<br>Batch Size=1<br>(ms) |
|-------------------------------|-----------|-------------------|--------------------------|
| ResNeXt101_<br>32x8d_wsl | 224 | 256 | 19.127 |
| ResNeXt101_<br>32x16d_wsl | 224 | 256 | 23.629 |
| ResNeXt101_<br>32x32d_wsl | 224 | 256 | 40.214 |
| ResNeXt101_<br>32x48d_wsl | 224 | 256 | 59.714 |
| Fix_ResNeXt101_<br>32x48d_wsl | 320 | 320 | 82.431 |
| EfficientNetB0 | 224 | 256 | 2.449 |
| EfficientNetB1 | 240 | 272 | 3.547 |
| EfficientNetB2 | 260 | 292 | 3.908 |
| EfficientNetB3 | 300 | 332 | 5.145 |
| EfficientNetB4 | 380 | 412 | 7.609 |
| EfficientNetB5 | 456 | 488 | 12.078 |
| EfficientNetB6 | 528 | 560 | 18.381 |
| EfficientNetB7 | 600 | 632 | 27.817 |
| EfficientNetB0_<br>small | 224 | 256 | 1.692 |
## Inference speed based on T4 GPU
| Models | Crop Size | Resize Short Size | FP16<br>Batch Size=1<br>(ms) | FP16<br>Batch Size=4<br>(ms) | FP16<br>Batch Size=8<br>(ms) | FP32<br>Batch Size=1<br>(ms) | FP32<br>Batch Size=4<br>(ms) | FP32<br>Batch Size=8<br>(ms) |
|---------------------------|-----------|-------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
| ResNeXt101_<br>32x8d_wsl | 224 | 256 | 18.19374 | 21.93529 | 34.67802 | 18.52528 | 34.25319 | 67.2283 |
| ResNeXt101_<br>32x16d_wsl | 224 | 256 | 18.52609 | 36.8288 | 62.79947 | 25.60395 | 71.88384 | 137.62327 |
| ResNeXt101_<br>32x32d_wsl | 224 | 256 | 33.51391 | 70.09682 | 125.81884 | 54.87396 | 160.04337 | 316.17718 |
| ResNeXt101_<br>32x48d_wsl | 224 | 256 | 50.97681 | 137.60926 | 190.82628 | 99.01698256 | 315.91261 | 551.83695 |
| Fix_ResNeXt101_<br>32x48d_wsl | 320 | 320 | 78.62869 | 191.76039 | 317.15436 | 160.0838242 | 595.99296 | 1151.47384 |
| EfficientNetB0 | 224 | 256 | 3.40122 | 5.95851 | 9.10801 | 3.442 | 6.11476 | 9.3304 |
| EfficientNetB1 | 240 | 272 | 5.25172 | 9.10233 | 14.11319 | 5.3322 | 9.41795 | 14.60388 |
| EfficientNetB2 | 260 | 292 | 5.91052 | 10.5898 | 17.38106 | 6.29351 | 10.95702 | 17.75308 |
| EfficientNetB3 | 300 | 332 | 7.69582 | 16.02548 | 27.4447 | 7.67749 | 16.53288 | 28.5939 |
| EfficientNetB4 | 380 | 412 | 11.55585 | 29.44261 | 53.97363 | 12.15894 | 30.94567 | 57.38511 |
| EfficientNetB5 | 456 | 488 | 19.63083 | 56.52299 | - | 20.48571 | 61.60252 | - |
| EfficientNetB6 | 528 | 560 | 30.05911 | - | - | 32.62402 | - | - |
| EfficientNetB7 | 600 | 632 | 47.86087 | - | - | 53.93823 | - | - |
| EfficientNetB0_small | 224 | 256 | 2.39166 | 4.36748 | 6.96002 | 2.3076 | 4.71886 | 7.21888 |
docs/en/models/HRNet_en.md 0 → 100644
浏览文件 @ f5168a7a
# HRNet series
## Overview
HRNet is a brand new neural network proposed by Microsoft research Asia in 2019. Different from the previous convolutional neural network, this network can still maintain high resolution in the deep layer of the network, so the heat map of the key points predicted is more accurate, and it is also more accurate in space. In addition, the network performs particularly well in other visual tasks sensitive to resolution, such as detection and segmentation.
The FLOPS, parameters, and inference time on the T4 GPU of this series of models are shown in the figure below.
![](../../images/models/T4_benchmark/t4.fp32.bs4.HRNet.flops.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.HRNet.params.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.HRNet.png)
![](../../images/models/T4_benchmark/t4.fp16.bs4.HRNet.png)
At present, there are 7 pretrained models of such models open-sourced by PaddleClas, and their indicators are shown in the figure. Among them, the reason why the accuracy of the HRNet_W48_C indicator is abnormal may be due to fluctuations in training.
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Parameters<br>(M) |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| HRNet_W18_C | 0.769 | 0.934 | 0.768 | 0.934 | 4.140 | 21.290 |
| HRNet_W30_C | 0.780 | 0.940 | 0.782 | 0.942 | 16.230 | 37.710 |
| HRNet_W32_C | 0.783 | 0.942 | 0.785 | 0.942 | 17.860 | 41.230 |
| HRNet_W40_C | 0.788 | 0.945 | 0.789 | 0.945 | 25.410 | 57.550 |
| HRNet_W44_C | 0.790 | 0.945 | 0.789 | 0.944 | 29.790 | 67.060 |
| HRNet_W48_C | 0.790 | 0.944 | 0.793 | 0.945 | 34.580 | 77.470 |
| HRNet_W64_C | 0.793 | 0.946 | 0.795 | 0.946 | 57.830 | 128.060 |
## Inference speed based on V100 GPU
| Models | Crop Size | Resize Short Size | FP32<br>Batch Size=1<br>(ms) |
|-------------|-----------|-------------------|--------------------------|
| HRNet_W18_C | 224 | 256 | 7.368 |
| HRNet_W30_C | 224 | 256 | 9.402 |
| HRNet_W32_C | 224 | 256 | 9.467 |
| HRNet_W40_C | 224 | 256 | 10.739 |
| HRNet_W44_C | 224 | 256 | 11.497 |
| HRNet_W48_C | 224 | 256 | 12.165 |
| HRNet_W64_C | 224 | 256 | 15.003 |
## Inference speed based on T4 GPU
| Models | Crop Size | Resize Short Size | FP16<br>Batch Size=1<br>(ms) | FP16<br>Batch Size=4<br>(ms) | FP16<br>Batch Size=8<br>(ms) | FP32<br>Batch Size=1<br>(ms) | FP32<br>Batch Size=4<br>(ms) | FP32<br>Batch Size=8<br>(ms) |
|-------------|-----------|-------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
| HRNet_W18_C | 224 | 256 | 6.79093 | 11.50986 | 17.67244 | 7.40636 | 13.29752 | 23.33445 |
| HRNet_W30_C | 224 | 256 | 8.98077 | 14.08082 | 21.23527 | 9.57594 | 17.35485 | 32.6933 |
| HRNet_W32_C | 224 | 256 | 8.82415 | 14.21462 | 21.19804 | 9.49807 | 17.72921 | 32.96305 |
| HRNet_W40_C | 224 | 256 | 11.4229 | 19.1595 | 30.47984 | 12.12202 | 25.68184 | 48.90623 |
| HRNet_W44_C | 224 | 256 | 12.25778 | 22.75456 | 32.61275 | 13.19858 | 32.25202 | 59.09871 |
| HRNet_W48_C | 224 | 256 | 12.65015 | 23.12886 | 33.37859 | 13.70761 | 34.43572 | 63.01219 |
| HRNet_W64_C | 224 | 256 | 15.10428 | 27.68901 | 40.4198 | 17.57527 | 47.9533 | 97.11228 |
docs/en/models/Inception_en.md 0 → 100644
浏览文件 @ f5168a7a
# Inception series
## Overview
GoogLeNet is a new neural network structure designed by Google in 2014, which, together with VGG network, became the twin champions of the ImageNet challenge that year. GoogLeNet introduces the Inception structure for the first time, and stacks the Inception structure in the network so that the number of network layers reaches 22, which is also the mark of the convolutional network exceeding 20 layers for the first time. Since 1x1 convolution is used in the Inception structure to reduce the dimension of channel number, and Global pooling is used to replace the traditional method of processing features in multiple fc layers, the final GoogLeNet network has much less FLOPS and parameters than VGG network, which has become a beautiful scenery of neural network design at that time.
Xception is another improvement to InceptionV3 that Google proposed after Inception. In Xception, the author used the depthwise separable convolution to replace the traditional convolution operation, which greatly saved the network FLOPS and the number of parameters, but improved the accuracy. In DeeplabV3+, the author further improved the Xception and increased the number of Xception layers, and designed the network of Xception65 and Xception71.
InceptionV4 is a new neural network designed by Google in 2016, when residual structure were all the rage, but the authors believe that high performance can be achieved using only Inception structure. InceptionV4 uses more Inception structure to achieve even greater precision on Imagenet-1k.
The FLOPS, parameters, and inference time on the T4 GPU of this series of models are shown in the figure below.
![](../../images/models/T4_benchmark/t4.fp32.bs4.Inception.flops.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.Inception.params.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.Inception.png)
![](../../images/models/T4_benchmark/t4.fp16.bs4.Inception.png)
The figure above reflects the relationship between the accuracy of Xception series and InceptionV4 and other indicators. Among them, Xception_deeplab is consistent with the structure of the paper, and Xception is an improved model developed by PaddleClas, which improves the accuracy by about 0.6% when the inference speed is basically unchanged. Details of the improved model are being updated, so stay tuned.
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Parameters<br>(M) |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| GoogLeNet | 0.707 | 0.897 | 0.698 | | 2.880 | 8.460 |
| Xception41 | 0.793 | 0.945 | 0.790 | 0.945 | 16.740 | 22.690 |
| Xception41<br>_deeplab | 0.796 | 0.944 | | | 18.160 | 26.730 |
| Xception65 | 0.810 | 0.955 | | | 25.950 | 35.480 |
| Xception65<br>_deeplab | 0.803 | 0.945 | | | 27.370 | 39.520 |
| Xception71 | 0.811 | 0.955 | | | 31.770 | 37.280 |
| InceptionV4 | 0.808 | 0.953 | 0.800 | 0.950 | 24.570 | 42.680 |
## Inference speed based on V100 GPU
| Models | Crop Size | Resize Short Size | FP32<br>Batch Size=1<br>(ms) |
|------------------------|-----------|-------------------|--------------------------|
| GoogLeNet | 224 | 256 | 1.807 |
| Xception41 | 299 | 320 | 3.972 |
| Xception41_<br>deeplab | 299 | 320 | 4.408 |
| Xception65 | 299 | 320 | 6.174 |
| Xception65_<br>deeplab | 299 | 320 | 6.464 |
| Xception71 | 299 | 320 | 6.782 |
| InceptionV4 | 299 | 320 | 11.141 |
## Inference speed based on T4 GPU
| Models | Crop Size | Resize Short Size | FP16<br>Batch Size=1<br>(ms) | FP16<br>Batch Size=4<br>(ms) | FP16<br>Batch Size=8<br>(ms) | FP32<br>Batch Size=1<br>(ms) | FP32<br>Batch Size=4<br>(ms) | FP32<br>Batch Size=8<br>(ms) |
|--------------------|-----------|-------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
| GoogLeNet | 299 | 320 | 1.75451 | 3.39931 | 4.71909 | 1.88038 | 4.48882 | 6.94035 |
| Xception41 | 299 | 320 | 2.91192 | 7.86878 | 15.53685 | 4.96939 | 17.01361 | 32.67831 |
| Xception41_<br>deeplab | 299 | 320 | 2.85934 | 7.2075 | 14.01406 | 5.33541 | 17.55938 | 33.76232 |
| Xception65 | 299 | 320 | 4.30126 | 11.58371 | 23.22213 | 7.26158 | 25.88778 | 53.45426 |
| Xception65_<br>deeplab | 299 | 320 | 4.06803 | 9.72694 | 19.477 | 7.60208 | 26.03699 | 54.74724 |
| Xception71 | 299 | 320 | 4.80889 | 13.5624 | 27.18822 | 8.72457 | 31.55549 | 69.31018 |
| InceptionV4 | 299 | 320 | 9.50821 | 13.72104 | 20.27447 | 12.99342 | 25.23416 | 43.56121 |
docs/en/models/Mobile_en.md 0 → 100644
浏览文件 @ f5168a7a
# Mobile and Embedded Vision Applications Network series
## Overview
MobileNetV1 is a network launched by Google in 2017 for use on mobile devices or embedded devices. The network replaces the depthwise separable convolution with the traditional convolution operation, that is, the combination of depthwise convolution and pointwise convolution. Compared with the traditional convolution operation, this combination can greatly save the number of parameters and computation. At the same time, MobileNetV1 can also be used for object detection, image segmentation and other visual tasks.
MobileNetV2 is a lightweight network proposed by Google following MobileNetV1. Compared with MobileNetV1, MobileNetV2 proposed Linear bottlenecks and Inverted residual block as a basic network structures, to constitute MobileNetV2 network architecture through stacking these basic module a lot. In the end, higher classification accuracy was achieved when FLOPS was only half of MobileNetV1.
The ShuffleNet series network is the lightweight network structure proposed by MEGVII. So far, there are two typical structures in this series network, namely, ShuffleNetV1 and ShuffleNetV2. A Channel Shuffle operation in ShuffleNet can exchange information between groups and perform end-to-end training. In the paper of ShuffleNetV2, the author proposes four criteria for designing lightweight networks, and designs the ShuffleNetV2 network according to the four criteria and the shortcomings of ShuffleNetV1.
MobileNetV3 is a new and lightweight network based on NAS proposed by Google in 2019. In order to further improve the effect, the activation functions of relu and sigmoid were replaced with hard_swish and hard_sigmoid activation functions, and some improved strategies were introduced to reduce the amount of network computing.
![](../../images/models/mobile_arm_top1.png)
![](../../images/models/mobile_arm_storage.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.mobile_trt.flops.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.mobile_trt.params.png)
Currently there are 32 pretrained models of the mobile series open source by PaddleClas, and their indicators are shown in the figure below. As you can see from the picture, newer lightweight models tend to perform better, and MobileNetV3 represents the latest lightweight neural network architecture. In MobileNetV3, the author used 1x1 convolution after global-avg-pooling in order to obtain higher accuracy,this operation significantly increases the number of parameters but has little impact on the amount of computation, so if the model is evaluated from a storage perspective of excellence, MobileNetV3 does not have much advantage, but because of its smaller computation, it has a faster inference speed. In addition, the SSLD distillation model in our model library performs excellently, refreshing the accuracy of the current lightweight model from various perspectives. Due to the complex structure and many branches of the MobileNetV3 model, which is not GPU friendly, the GPU inference speed is not as good as that of MobileNetV1.
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Parameters<br>(M) |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| MobileNetV1_x0_25 | 0.514 | 0.755 | 0.506 | | 0.070 | 0.460 |
| MobileNetV1_x0_5 | 0.635 | 0.847 | 0.637 | | 0.280 | 1.310 |
| MobileNetV1_x0_75 | 0.688 | 0.882 | 0.684 | | 0.630 | 2.550 |
| MobileNetV1 | 0.710 | 0.897 | 0.706 | | 1.110 | 4.190 |
| MobileNetV1_ssld | 0.779 | 0.939 | | | 1.110 | 4.190 |
| MobileNetV2_x0_25 | 0.532 | 0.765 | | | 0.050 | 1.500 |
| MobileNetV2_x0_5 | 0.650 | 0.857 | 0.654 | 0.864 | 0.170 | 1.930 |
| MobileNetV2_x0_75 | 0.698 | 0.890 | 0.698 | 0.896 | 0.350 | 2.580 |
| MobileNetV2 | 0.722 | 0.907 | 0.718 | 0.910 | 0.600 | 3.440 |
| MobileNetV2_x1_5 | 0.741 | 0.917 | | | 1.320 | 6.760 |
| MobileNetV2_x2_0 | 0.752 | 0.926 | | | 2.320 | 11.130 |
| MobileNetV2_ssld | 0.7674 | 0.9339 | | | 0.600 | 3.440 |
| MobileNetV3_large_<br>x1_25 | 0.764 | 0.930 | 0.766 | | 0.714 | 7.440 |
| MobileNetV3_large_<br>x1_0 | 0.753 | 0.923 | 0.752 | | 0.450 | 5.470 |
| MobileNetV3_large_<br>x0_75 | 0.731 | 0.911 | 0.733 | | 0.296 | 3.910 |
| MobileNetV3_large_<br>x0_5 | 0.692 | 0.885 | 0.688 | | 0.138 | 2.670 |
| MobileNetV3_large_<br>x0_35 | 0.643 | 0.855 | 0.642 | | 0.077 | 2.100 |
| MobileNetV3_small_<br>x1_25 | 0.707 | 0.895 | 0.704 | | 0.195 | 3.620 |
| MobileNetV3_small_<br>x1_0 | 0.682 | 0.881 | 0.675 | | 0.123 | 2.940 |
| MobileNetV3_small_<br>x0_75 | 0.660 | 0.863 | 0.654 | | 0.088 | 2.370 |
| MobileNetV3_small_<br>x0_5 | 0.592 | 0.815 | 0.580 | | 0.043 | 1.900 |
| MobileNetV3_small_<br>x0_35 | 0.530 | 0.764 | 0.498 | | 0.026 | 1.660 |
| MobileNetV3_large_<br>x1_0_ssld | 0.790 | 0.945 | | | 0.450 | 5.470 |
| MobileNetV3_large_<br>x1_0_ssld_int8 | 0.761 | | | | | |
| MobileNetV3_small_<br>x1_0_ssld | 0.713 | 0.901 | | | 0.123 | 2.940 |
| ShuffleNetV2 | 0.688 | 0.885 | 0.694 | | 0.280 | 2.260 |
| ShuffleNetV2_x0_25 | 0.499 | 0.738 | | | 0.030 | 0.600 |
| ShuffleNetV2_x0_33 | 0.537 | 0.771 | | | 0.040 | 0.640 |
| ShuffleNetV2_x0_5 | 0.603 | 0.823 | 0.603 | | 0.080 | 1.360 |
| ShuffleNetV2_x1_5 | 0.716 | 0.902 | 0.726 | | 0.580 | 3.470 |
| ShuffleNetV2_x2_0 | 0.732 | 0.912 | 0.749 | | 1.120 | 7.320 |
| ShuffleNetV2_swish | 0.700 | 0.892 | | | 0.290 | 2.260 |
## Inference speed and storage size based on SD855
| Models | Batch Size=1(ms) | Storage Size(M) |
|:--:|:--:|:--:|
| MobileNetV1_x0_25 | 3.220 | 1.900 |
| MobileNetV1_x0_5 | 9.580 | 5.200 |
| MobileNetV1_x0_75 | 19.436 | 10.000 |
| MobileNetV1 | 32.523 | 16.000 |
| MobileNetV1_ssld | 32.523 | 16.000 |
| MobileNetV2_x0_25 | 3.799 | 6.100 |
| MobileNetV2_x0_5 | 8.702 | 7.800 |
| MobileNetV2_x0_75 | 15.531 | 10.000 |
| MobileNetV2 | 23.318 | 14.000 |
| MobileNetV2_x1_5 | 45.624 | 26.000 |
| MobileNetV2_x2_0 | 74.292 | 43.000 |
| MobileNetV2_ssld | 23.318 | 14.000 |
| MobileNetV3_large_x1_25 | 28.218 | 29.000 |
| MobileNetV3_large_x1_0 | 19.308 | 21.000 |
| MobileNetV3_large_x0_75 | 13.565 | 16.000 |
| MobileNetV3_large_x0_5 | 7.493 | 11.000 |
| MobileNetV3_large_x0_35 | 5.137 | 8.600 |
| MobileNetV3_small_x1_25 | 9.275 | 14.000 |
| MobileNetV3_small_x1_0 | 6.546 | 12.000 |
| MobileNetV3_small_x0_75 | 5.284 | 9.600 |
| MobileNetV3_small_x0_5 | 3.352 | 7.800 |
| MobileNetV3_small_x0_35 | 2.635 | 6.900 |
| MobileNetV3_large_x1_0_ssld | 19.308 | 21.000 |
| MobileNetV3_large_x1_0_ssld_int8 | 14.395 | 10.000 |
| MobileNetV3_small_x1_0_ssld | 6.546 | 12.000 |
| ShuffleNetV2 | 10.941 | 9.000 |
| ShuffleNetV2_x0_25 | 2.329 | 2.700 |
| ShuffleNetV2_x0_33 | 2.643 | 2.800 |
| ShuffleNetV2_x0_5 | 4.261 | 5.600 |
| ShuffleNetV2_x1_5 | 19.352 | 14.000 |
| ShuffleNetV2_x2_0 | 34.770 | 28.000 |
| ShuffleNetV2_swish | 16.023 | 9.100 |
## Inference speed based on T4 GPU
| Models | FP16<br>Batch Size=1<br>(ms) | FP16<br>Batch Size=4<br>(ms) | FP16<br>Batch Size=8<br>(ms) | FP32<br>Batch Size=1<br>(ms) | FP32<br>Batch Size=4<br>(ms) | FP32<br>Batch Size=8<br>(ms) |
|-----------------------------|-----------------------|-----------------------|-----------------------|-----------------------|-----------------------|-----------------------|
| MobileNetV1_x0_25 | 0.68422 | 1.13021 | 1.72095 | 0.67274 | 1.226 | 1.84096 |
| MobileNetV1_x0_5 | 0.69326 | 1.09027 | 1.84746 | 0.69947 | 1.43045 | 2.39353 |
| MobileNetV1_x0_75 | 0.6793 | 1.29524 | 2.15495 | 0.79844 | 1.86205 | 3.064 |
| MobileNetV1 | 0.71942 | 1.45018 | 2.47953 | 0.91164 | 2.26871 | 3.90797 |
| MobileNetV1_ssld | 0.71942 | 1.45018 | 2.47953 | 0.91164 | 2.26871 | 3.90797 |
| MobileNetV2_x0_25 | 2.85399 | 3.62405 | 4.29952 | 2.81989 | 3.52695 | 4.2432 |
| MobileNetV2_x0_5 | 2.84258 | 3.1511 | 4.10267 | 2.80264 | 3.65284 | 4.31737 |
| MobileNetV2_x0_75 | 2.82183 | 3.27622 | 4.98161 | 2.86538 | 3.55198 | 5.10678 |
| MobileNetV2 | 2.78603 | 3.71982 | 6.27879 | 2.62398 | 3.54429 | 6.41178 |
| MobileNetV2_x1_5 | 2.81852 | 4.87434 | 8.97934 | 2.79398 | 5.30149 | 9.30899 |
| MobileNetV2_x2_0 | 3.65197 | 6.32329 | 11.644 | 3.29788 | 7.08644 | 12.45375 |
| MobileNetV2_ssld | 2.78603 | 3.71982 | 6.27879 | 2.62398 | 3.54429 | 6.41178 |
| MobileNetV3_large_x1_25 | 2.34387 | 3.16103 | 4.79742 | 2.35117 | 3.44903 | 5.45658 |
| MobileNetV3_large_x1_0 | 2.20149 | 3.08423 | 4.07779 | 2.04296 | 2.9322 | 4.53184 |
| MobileNetV3_large_x0_75 | 2.1058 | 2.61426 | 3.61021 | 2.0006 | 2.56987 | 3.78005 |
| MobileNetV3_large_x0_5 | 2.06934 | 2.77341 | 3.35313 | 2.11199 | 2.88172 | 3.19029 |
| MobileNetV3_large_x0_35 | 2.14965 | 2.7868 | 3.36145 | 1.9041 | 2.62951 | 3.26036 |
| MobileNetV3_small_x1_25 | 2.06817 | 2.90193 | 3.5245 | 2.02916 | 2.91866 | 3.34528 |
| MobileNetV3_small_x1_0 | 1.73933 | 2.59478 | 3.40276 | 1.74527 | 2.63565 | 3.28124 |
| MobileNetV3_small_x0_75 | 1.80617 | 2.64646 | 3.24513 | 1.93697 | 2.64285 | 3.32797 |
| MobileNetV3_small_x0_5 | 1.95001 | 2.74014 | 3.39485 | 1.88406 | 2.99601 | 3.3908 |
| MobileNetV3_small_x0_35 | 2.10683 | 2.94267 | 3.44254 | 1.94427 | 2.94116 | 3.41082 |
| MobileNetV3_large_x1_0_ssld | 2.20149 | 3.08423 | 4.07779 | 2.04296 | 2.9322 | 4.53184 |
| MobileNetV3_small_x1_0_ssld | 1.73933 | 2.59478 | 3.40276 | 1.74527 | 2.63565 | 3.28124 |
| ShuffleNetV2 | 1.95064 | 2.15928 | 2.97169 | 1.89436 | 2.26339 | 3.17615 |
| ShuffleNetV2_x0_25 | 1.43242 | 2.38172 | 2.96768 | 1.48698 | 2.29085 | 2.90284 |
| ShuffleNetV2_x0_33 | 1.69008 | 2.65706 | 2.97373 | 1.75526 | 2.85557 | 3.09688 |
| ShuffleNetV2_x0_5 | 1.48073 | 2.28174 | 2.85436 | 1.59055 | 2.18708 | 3.09141 |
| ShuffleNetV2_x1_5 | 1.51054 | 2.4565 | 3.41738 | 1.45389 | 2.5203 | 3.99872 |
| ShuffleNetV2_x2_0 | 1.95616 | 2.44751 | 4.19173 | 2.15654 | 3.18247 | 5.46893 |
| ShuffleNetV2_swish | 2.50213 | 2.92881 | 3.474 | 2.5129 | 2.97422 | 3.69357 |
docs/en/models/Others_en.md 0 → 100644
浏览文件 @ f5168a7a
# Other networks
## Overview
In 2012, AlexNet network proposed by Alex et al. won the ImageNet competition by far surpassing the second place, and the convolutional neural network and even deep learning attracted wide attention. AlexNet used relu as the activation function of CNN to solve the gradient dispersion problem of sigmoid when the network is deep. During the training, Dropout was used to randomly lose a part of the neurons, avoiding the overfitting of the model. In the network, overlapping maximum pooling is used to replace the average pooling commonly used in CNN, which avoids the fuzzy effect of average pooling and improves the feature richness. In a sense, AlexNet has exploded the research and application of neural networks.
SqueezeNet achieved the same precision as AlexNet on Imagenet-1k, but only with 1/50 parameters. The core of the network is the Fire module, which used the convolution of 1x1 to achieve channel dimensionality reduction, thus greatly saving the number of parameters. The author created SqueezeNet by stacking a large number of Fire modules.
VGG is a convolutional neural network developed by researchers at Oxford University's Visual Geometry Group and DeepMind. The network explores the relationship between the depth of the convolutional neural network and its performance. By repeatedly stacking the small convolutional kernel of 3x3 and the maximum pooling layer of 2x2, the multi-layer convolutional neural network is successfully constructed and has achieved good convergence accuracy. In the end, VGG won the runner-up of ILSVRC 2014 classification and the champion of positioning.
DarkNet53 is designed for object detection by YOLO author in the paper. The network is basically composed of 1x1 and 3x3 kernel, with a total of 53 layers, named DarkNet53.
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Parameters<br>(M) |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| AlexNet | 0.567 | 0.792 | 0.5720 | | 1.370 | 61.090 |
| SqueezeNet1_0 | 0.596 | 0.817 | 0.575 | | 1.550 | 1.240 |
| SqueezeNet1_1 | 0.601 | 0.819 | | | 0.690 | 1.230 |
| VGG11 | 0.693 | 0.891 | | | 15.090 | 132.850 |
| VGG13 | 0.700 | 0.894 | | | 22.480 | 133.030 |
| VGG16 | 0.720 | 0.907 | 0.715 | 0.901 | 30.810 | 138.340 |
| VGG19 | 0.726 | 0.909 | | | 39.130 | 143.650 |
| DarkNet53 | 0.780 | 0.941 | 0.772 | 0.938 | 18.580 | 41.600 |
| ResNet50_ACNet | 0.767 | 0.932 | | | 10.730 | 33.110 |
| ResNet50_ACNet<br>_deploy | 0.767 | 0.932 | | | 8.190 | 25.550 |
## Inference speed based on V100 GPU
| Models | Crop Size | Resize Short Size | FP32<br>Batch Size=1<br>(ms) |
|---------------------------|-----------|-------------------|----------------------|
| AlexNet | 224 | 256 | 1.176 |
| SqueezeNet1_0 | 224 | 256 | 0.860 |
| SqueezeNet1_1 | 224 | 256 | 0.763 |
| VGG11 | 224 | 256 | 1.867 |
| VGG13 | 224 | 256 | 2.148 |
| VGG16 | 224 | 256 | 2.616 |
| VGG19 | 224 | 256 | 3.076 |
| DarkNet53 | 256 | 256 | 3.139 |
| ResNet50_ACNet<br>_deploy | 224 | 256 | 5.626 |
## Inference speed based on T4 GPU
| Models | Crop Size | Resize Short Size | FP16<br>Batch Size=1<br>(ms) | FP16<br>Batch Size=4<br>(ms) | FP16<br>Batch Size=8<br>(ms) | FP32<br>Batch Size=1<br>(ms) | FP32<br>Batch Size=4<br>(ms) | FP32<br>Batch Size=8<br>(ms) |
|-----------------------|-----------|-------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
| AlexNet | 224 | 256 | 1.06447 | 1.70435 | 2.38402 | 1.44993 | 2.46696 | 3.72085 |
| SqueezeNet1_0 | 224 | 256 | 0.97162 | 2.06719 | 3.67499 | 0.96736 | 2.53221 | 4.54047 |
| SqueezeNet1_1 | 224 | 256 | 0.81378 | 1.62919 | 2.68044 | 0.76032 | 1.877 | 3.15298 |
| VGG11 | 224 | 256 | 2.24408 | 4.67794 | 7.6568 | 3.90412 | 9.51147 | 17.14168 |
| VGG13 | 224 | 256 | 2.58589 | 5.82708 | 10.03591 | 4.64684 | 12.61558 | 23.70015 |
| VGG16 | 224 | 256 | 3.13237 | 7.19257 | 12.50913 | 5.61769 | 16.40064 | 32.03939 |
| VGG19 | 224 | 256 | 3.69987 | 8.59168 | 15.07866 | 6.65221 | 20.4334 | 41.55902 |
| DarkNet53 | 256 | 256 | 3.18101 | 5.88419 | 10.14964 | 4.10829 | 12.1714 | 22.15266 |
| ResNet50_ACNet | 256 | 256 | 3.89002 | 4.58195 | 9.01095 | 5.33395 | 10.96843 | 18.70368 |
| ResNet50_ACNet_deploy | 224 | 256 | 2.6823 | 5.944 | 7.16655 | 3.49161 | 7.78374 | 13.94361 |
docs/en/models/ResNet_and_vd_en.md 0 → 100644
浏览文件 @ f5168a7a
# ResNet and ResNet_vd series
## Overview
The ResNet series model was proposed in 2015 and won the championship in the ILSVRC2015 competition with a top5 error rate of 3.57%. The network innovatively proposed the residual structure, and built the ResNet network by stacking multiple residual structures. Experiments show that using residual blocks can improve the convergence speed and accuracy effectively.
Joyce Xu of Stanford university calls ResNet one of three architectures that "really redefine the way we think about neural networks." Due to the outstanding performance of ResNet, more and more scholars and engineers from academia and industry have improved its structure. The well-known ones include wide-resnet, resnet-vc, resnet-vd, Res2Net, etc. The number of parameters and FLOPs of resnet-vc and resnet-vd are almost the same as those of ResNet, so we hereby unified them into the ResNet series.
The models of the ResNet series released this time include 14 pre-trained models including ResNet50, ResNet50_vd, ResNet50_vd_ssld, and ResNet200_vd. At the training level, ResNet adopted the standard training process for training ImageNet, while the rest of the improved model adopted more training strategies, such as cosine decay for the decline of learning rate and the regular label smoothing method,mixup was added to the data preprocessing, and the total number of iterations increased from 120 epoches to 200 epoches.
Among them, ResNet50_vd_v2 and ResNet50_vd_ssld adopted knowledge distillation, which further improved the accuracy of the model while keeping the structure unchanged. Specifically, the teacher model of ResNet50_vd_v2 is ResNet152_vd (top1 accuracy 80.59%), the training set is imagenet-1k, the teacher model of ResNet50_vd_ssld is ResNeXt101_32x16d_wsl (top1 accuracy 84.2%), and the training set is the combination of 4 million data mined by imagenet-22k and ImageNet-1k . The specific methods of knowledge distillation are being continuously updated.
The FLOPS, parameters, and inference time on the T4 GPU of this series of models are shown in the figure below.
![](../../images/models/T4_benchmark/t4.fp32.bs4.ResNet.flops.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.ResNet.params.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.ResNet.png)
![](../../images/models/T4_benchmark/t4.fp16.bs4.ResNet.png)
As can be seen from the above curves, the higher the number of layers, the higher the accuracy, but the corresponding number of parameters, calculation and latency will increase. ResNet50_vd_ssld further improves the accuracy of top-1 of the ImageNet-1k validation set by using stronger teachers and more data, reaching 82.39%, refreshing the accuracy of ResNet50 series models.
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Parameters<br>(M) |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| ResNet18 | 0.710 | 0.899 | 0.696 | 0.891 | 3.660 | 11.690 |
| ResNet18_vd | 0.723 | 0.908 | | | 4.140 | 11.710 |
| ResNet34 | 0.746 | 0.921 | 0.732 | 0.913 | 7.360 | 21.800 |
| ResNet34_vd | 0.760 | 0.930 | | | 7.390 | 21.820 |
| ResNet50 | 0.765 | 0.930 | 0.760 | 0.930 | 8.190 | 25.560 |
| ResNet50_vc | 0.784 | 0.940 | | | 8.670 | 25.580 |
| ResNet50_vd | 0.791 | 0.944 | 0.792 | 0.946 | 8.670 | 25.580 |
| ResNet50_vd_v2 | 0.798 | 0.949 | | | 8.670 | 25.580 |
| ResNet101 | 0.776 | 0.936 | 0.776 | 0.938 | 15.520 | 44.550 |
| ResNet101_vd | 0.802 | 0.950 | | | 16.100 | 44.570 |
| ResNet152 | 0.783 | 0.940 | 0.778 | 0.938 | 23.050 | 60.190 |
| ResNet152_vd | 0.806 | 0.953 | | | 23.530 | 60.210 |
| ResNet200_vd | 0.809 | 0.953 | | | 30.530 | 74.740 |
| ResNet50_vd_ssld | 0.824 | 0.961 | | | 8.670 | 25.580 |
| ResNet50_vd_ssld_v2 | 0.830 | 0.964 | | | 8.670 | 25.580 |
| Fix_ResNet50_vd_ssld_v2 | 0.840 | 0.970 | | | 17.696 | 25.580 |
| ResNet101_vd_ssld | 0.837 | 0.967 | | | 16.100 | 44.570 |
* Note: `ResNet50_vd_ssld_v2` is obtained by adding AutoAugment in training process on the basis of `ResNet50_vd_ssld` training strategy.`Fix_ResNet50_vd_ssld_v2` stopped all parameter updates of `ResNet50_vd_ssld_v2` except the FC layer,and fine-tuned on ImageNet1k dataset, the resolution is 320x320.
## Inference speed based on V100 GPU
| Models | Crop Size | Resize Short Size | FP32<br>Batch Size=1<br>(ms) |
|------------------|-----------|-------------------|--------------------------|
| ResNet18 | 224 | 256 | 1.499 |
| ResNet18_vd | 224 | 256 | 1.603 |
| ResNet34 | 224 | 256 | 2.272 |
| ResNet34_vd | 224 | 256 | 2.343 |
| ResNet50 | 224 | 256 | 2.939 |
| ResNet50_vc | 224 | 256 | 3.041 |
| ResNet50_vd | 224 | 256 | 3.165 |
| ResNet50_vd_v2 | 224 | 256 | 3.165 |
| ResNet101 | 224 | 256 | 5.314 |
| ResNet101_vd | 224 | 256 | 5.252 |
| ResNet152 | 224 | 256 | 7.205 |
| ResNet152_vd | 224 | 256 | 7.200 |
| ResNet200_vd | 224 | 256 | 8.885 |
| ResNet50_vd_ssld | 224 | 256 | 3.165 |
| ResNet101_vd_ssld | 224 | 256 | 5.252 |
## Inference speed based on T4 GPU
| Models | Crop Size | Resize Short Size | FP16<br>Batch Size=1<br>(ms) | FP16<br>Batch Size=4<br>(ms) | FP16<br>Batch Size=8<br>(ms) | FP32<br>Batch Size=1<br>(ms) | FP32<br>Batch Size=4<br>(ms) | FP32<br>Batch Size=8<br>(ms) |
|-------------------|-----------|-------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
| ResNet18 | 224 | 256 | 1.3568 | 2.5225 | 3.61904 | 1.45606 | 3.56305 | 6.28798 |
| ResNet18_vd | 224 | 256 | 1.39593 | 2.69063 | 3.88267 | 1.54557 | 3.85363 | 6.88121 |
| ResNet34 | 224 | 256 | 2.23092 | 4.10205 | 5.54904 | 2.34957 | 5.89821 | 10.73451 |
| ResNet34_vd | 224 | 256 | 2.23992 | 4.22246 | 5.79534 | 2.43427 | 6.22257 | 11.44906 |
| ResNet50 | 224 | 256 | 2.63824 | 4.63802 | 7.02444 | 3.47712 | 7.84421 | 13.90633 |
| ResNet50_vc | 224 | 256 | 2.67064 | 4.72372 | 7.17204 | 3.52346 | 8.10725 | 14.45577 |
| ResNet50_vd | 224 | 256 | 2.65164 | 4.84109 | 7.46225 | 3.53131 | 8.09057 | 14.45965 |
| ResNet50_vd_v2 | 224 | 256 | 2.65164 | 4.84109 | 7.46225 | 3.53131 | 8.09057 | 14.45965 |
| ResNet101 | 224 | 256 | 5.04037 | 7.73673 | 10.8936 | 6.07125 | 13.40573 | 24.3597 |
| ResNet101_vd | 224 | 256 | 5.05972 | 7.83685 | 11.34235 | 6.11704 | 13.76222 | 25.11071 |
| ResNet152 | 224 | 256 | 7.28665 | 10.62001 | 14.90317 | 8.50198 | 19.17073 | 35.78384 |
| ResNet152_vd | 224 | 256 | 7.29127 | 10.86137 | 15.32444 | 8.54376 | 19.52157 | 36.64445 |
| ResNet200_vd | 224 | 256 | 9.36026 | 13.5474 | 19.0725 | 10.80619 | 25.01731 | 48.81399 |
| ResNet50_vd_ssld | 224 | 256 | 2.65164 | 4.84109 | 7.46225 | 3.53131 | 8.09057 | 14.45965 |
| ResNet50_vd_ssld_v2 | 224 | 256 | 2.65164 | 4.84109 | 7.46225 | 3.53131 | 8.09057 | 14.45965 |
| Fix_ResNet50_vd_ssld_v2 | 320 | 320 | 3.42818 | 7.51534 | 13.19370 | 5.07696 | 14.64218 | 27.01453 |
| ResNet101_vd_ssld | 224 | 256 | 5.05972 | 7.83685 | 11.34235 | 6.11704 | 13.76222 | 25.11071 |
docs/en/models/SEResNext_and_Res2Net_en.md 0 → 100644
浏览文件 @ f5168a7a
# SEResNeXt and Res2Net series
## Overview
ResNeXt, one of the typical variants of ResNet, was presented at the CVPR conference in 2017. Prior to this, the methods to improve the model accuracy mainly focused on deepening or widening the network, which increased the number of parameters and calculation, and slowed down the inference speed accordingly. The concept of cardinality was proposed in ResNeXt structure. The author found that increasing the number of channel groups was more effective than increasing the depth and width through experiments. It can improve the accuracy without increasing the parameter complexity and reduce the number of parameters at the same time, so it is a more successful variant of ResNet.
SENet is the winner of the 2017 ImageNet classification competition. It proposes a new SE structure that can be migrated to any other network. It controls the scale to enhance the important features between each channel, and weaken the unimportant features. So that the extracted features are more directional.
Res2Net is a brand-new improvement of ResNet proposed in 2019. The solution can be easily integrated with other excellent modules. Without increasing the amount of calculation, the performance on ImageNet, CIFAR-100 and other data sets exceeds ResNet. Res2Net, with its simple structure and superior performance, further explores the multi-scale representation capability of CNN at a more fine-grained level. Res2Net reveals a new dimension to improve model accuracy, called scale, which is an essential and more effective factor in addition to the existing dimensions of depth, width, and cardinality. The network also performs well in other visual tasks such as object detection and image segmentation.
The FLOPS, parameters, and inference time on the T4 GPU of this series of models are shown in the figure below.
![](../../images/models/T4_benchmark/t4.fp32.bs4.SeResNeXt.flops.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.SeResNeXt.params.png)
![](../../images/models/T4_benchmark/t4.fp32.bs4.SeResNeXt.png)
![](../../images/models/T4_benchmark/t4.fp16.bs4.SeResNeXt.png)
At present, there are a total of 24 pretrained models of the three categories open sourced by PaddleClas, and the indicators are shown in the figure. It can be seen from the diagram that under the same Flops and Params, the improved model tends to have higher accuracy, but the inference speed is often inferior to the ResNet series. On the other hand, Res2Net performed better. Compared with group operation in ResNeXt and SE structure operation in SEResNet, Res2Net tended to have better accuracy in the same Flops, Params and inference speed.
## Accuracy, FLOPS and Parameters
| Models | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Parameters<br>(M) |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| Res2Net50_26w_4s | 0.793 | 0.946 | 0.780 | 0.936 | 8.520 | 25.700 |
| Res2Net50_vd_26w_4s | 0.798 | 0.949 | | | 8.370 | 25.060 |
| Res2Net50_14w_8s | 0.795 | 0.947 | 0.781 | 0.939 | 9.010 | 25.720 |
| Res2Net101_vd_26w_4s | 0.806 | 0.952 | | | 16.670 | 45.220 |
| Res2Net200_vd_26w_4s | 0.812 | 0.957 | | | 31.490 | 76.210 |
| ResNeXt50_32x4d | 0.778 | 0.938 | 0.778 | | 8.020 | 23.640 |
| ResNeXt50_vd_32x4d | 0.796 | 0.946 | | | 8.500 | 23.660 |
| ResNeXt50_64x4d | 0.784 | 0.941 | | | 15.060 | 42.360 |
| ResNeXt50_vd_64x4d | 0.801 | 0.949 | | | 15.540 | 42.380 |
| ResNeXt101_32x4d | 0.787 | 0.942 | 0.788 | | 15.010 | 41.540 |
| ResNeXt101_vd_32x4d | 0.803 | 0.951 | | | 15.490 | 41.560 |
| ResNeXt101_64x4d | 0.784 | 0.945 | 0.796 | | 29.050 | 78.120 |
| ResNeXt101_vd_64x4d | 0.808 | 0.952 | | | 29.530 | 78.140 |
| ResNeXt152_32x4d | 0.790 | 0.943 | | | 22.010 | 56.280 |
| ResNeXt152_vd_32x4d | 0.807 | 0.952 | | | 22.490 | 56.300 |
| ResNeXt152_64x4d | 0.795 | 0.947 | | | 43.030 | 107.570 |
| ResNeXt152_vd_64x4d | 0.811 | 0.953 | | | 43.520 | 107.590 |
| SE_ResNet18_vd | 0.733 | 0.914 | | | 4.140 | 11.800 |
| SE_ResNet34_vd | 0.765 | 0.932 | | | 7.840 | 21.980 |
| SE_ResNet50_vd | 0.795 | 0.948 | | | 8.670 | 28.090 |
| SE_ResNeXt50_32x4d | 0.784 | 0.940 | 0.789 | 0.945 | 8.020 | 26.160 |
| SE_ResNeXt50_vd_32x4d | 0.802 | 0.949 | | | 10.760 | 26.280 |
| SE_ResNeXt101_32x4d | 0.791 | 0.942 | 0.793 | 0.950 | 15.020 | 46.280 |
| SENet154_vd | 0.814 | 0.955 | | | 45.830 | 114.290 |
## Inference speed based on V100 GPU
| Models | Crop Size | Resize Short Size | FP32<br>Batch Size=1<br>(ms) |
|-----------------------|-----------|-------------------|--------------------------|
| Res2Net50_26w_4s | 224 | 256 | 4.148 |
| Res2Net50_vd_26w_4s | 224 | 256 | 4.172 |
| Res2Net50_14w_8s | 224 | 256 | 5.113 |
| Res2Net101_vd_26w_4s | 224 | 256 | 7.327 |
| Res2Net200_vd_26w_4s | 224 | 256 | 12.806 |
| ResNeXt50_32x4d | 224 | 256 | 10.964 |
| ResNeXt50_vd_32x4d | 224 | 256 | 7.566 |
| ResNeXt50_64x4d | 224 | 256 | 13.905 |
| ResNeXt50_vd_64x4d | 224 | 256 | 14.321 |
| ResNeXt101_32x4d | 224 | 256 | 14.915 |
| ResNeXt101_vd_32x4d | 224 | 256 | 14.885 |
| ResNeXt101_64x4d | 224 | 256 | 28.716 |
| ResNeXt101_vd_64x4d | 224 | 256 | 28.398 |
| ResNeXt152_32x4d | 224 | 256 | 22.996 |
| ResNeXt152_vd_32x4d | 224 | 256 | 22.729 |
| ResNeXt152_64x4d | 224 | 256 | 46.705 |
| ResNeXt152_vd_64x4d | 224 | 256 | 46.395 |
| SE_ResNet18_vd | 224 | 256 | 1.694 |
| SE_ResNet34_vd | 224 | 256 | 2.786 |
| SE_ResNet50_vd | 224 | 256 | 3.749 |
| SE_ResNeXt50_32x4d | 224 | 256 | 8.924 |
| SE_ResNeXt50_vd_32x4d | 224 | 256 | 9.011 |
| SE_ResNeXt101_32x4d | 224 | 256 | 19.204 |
| SENet154_vd | 224 | 256 | 50.406 |
## Inference speed based on T4 GPU
| Models | Crop Size | Resize Short Size | FP16<br>Batch Size=1<br>(ms) | FP16<br>Batch Size=4<br>(ms) | FP16<br>Batch Size=8<br>(ms) | FP32<br>Batch Size=1<br>(ms) | FP32<br>Batch Size=4<br>(ms) | FP32<br>Batch Size=8<br>(ms) |
|-----------------------|-----------|-------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|------------------------------|
| Res2Net50_26w_4s | 224 | 256 | 3.56067 | 6.61827 | 11.41566 | 4.47188 | 9.65722 | 17.54535 |
| Res2Net50_vd_26w_4s | 224 | 256 | 3.69221 | 6.94419 | 11.92441 | 4.52712 | 9.93247 | 18.16928 |
| Res2Net50_14w_8s | 224 | 256 | 4.45745 | 7.69847 | 12.30935 | 5.4026 | 10.60273 | 18.01234 |
| Res2Net101_vd_26w_4s | 224 | 256 | 6.53122 | 10.81895 | 18.94395 | 8.08729 | 17.31208 | 31.95762 |
| Res2Net200_vd_26w_4s | 224 | 256 | 11.66671 | 18.93953 | 33.19188 | 14.67806 | 32.35032 | 63.65899 |
| ResNeXt50_32x4d | 224 | 256 | 7.61087 | 8.88918 | 12.99674 | 7.56327 | 10.6134 | 18.46915 |
| ResNeXt50_vd_32x4d | 224 | 256 | 7.69065 | 8.94014 | 13.4088 | 7.62044 | 11.03385 | 19.15339 |
| ResNeXt50_64x4d | 224 | 256 | 13.78688 | 15.84655 | 21.79537 | 13.80962 | 18.4712 | 33.49843 |
| ResNeXt50_vd_64x4d | 224 | 256 | 13.79538 | 15.22201 | 22.27045 | 13.94449 | 18.88759 | 34.28889 |
| ResNeXt101_32x4d | 224 | 256 | 16.59777 | 17.93153 | 21.36541 | 16.21503 | 19.96568 | 33.76831 |
| ResNeXt101_vd_32x4d | 224 | 256 | 16.36909 | 17.45681 | 22.10216 | 16.28103 | 20.25611 | 34.37152 |
| ResNeXt101_64x4d | 224 | 256 | 30.12355 | 32.46823 | 38.41901 | 30.4788 | 36.29801 | 68.85559 |
| ResNeXt101_vd_64x4d | 224 | 256 | 30.34022 | 32.27869 | 38.72523 | 30.40456 | 36.77324 | 69.66021 |
| ResNeXt152_32x4d | 224 | 256 | 25.26417 | 26.57001 | 30.67834 | 24.86299 | 29.36764 | 52.09426 |
| ResNeXt152_vd_32x4d | 224 | 256 | 25.11196 | 26.70515 | 31.72636 | 25.03258 | 30.08987 | 52.64429 |
| ResNeXt152_64x4d | 224 | 256 | 46.58293 | 48.34563 | 56.97961 | 46.7564 | 56.34108 | 106.11736 |
| ResNeXt152_vd_64x4d | 224 | 256 | 47.68447 | 48.91406 | 57.29329 | 47.18638 | 57.16257 | 107.26288 |
| SE_ResNet18_vd | 224 | 256 | 1.61823 | 3.1391 | 4.60282 | 1.7691 | 4.19877 | 7.5331 |
| SE_ResNet34_vd | 224 | 256 | 2.67518 | 5.04694 | 7.18946 | 2.88559 | 7.03291 | 12.73502 |
| SE_ResNet50_vd | 224 | 256 | 3.65394 | 7.568 | 12.52793 | 4.28393 | 10.38846 | 18.33154 |
| SE_ResNeXt50_32x4d | 224 | 256 | 9.06957 | 11.37898 | 18.86282 | 8.74121 | 13.563 | 23.01954 |
| SE_ResNeXt50_vd_32x4d | 224 | 256 | 9.25016 | 11.85045 | 25.57004 | 9.17134 | 14.76192 | 19.914 |
| SE_ResNeXt101_32x4d | 224 | 256 | 19.34455 | 20.6104 | 32.20432 | 18.82604 | 25.31814 | 41.97758 |
| SENet154_vd | 224 | 256 | 49.85733 | 54.37267 | 74.70447 | 53.79794 | 66.31684 | 121.59885 |
docs/en/models/Tricks_en.md 0 → 100644
浏览文件 @ f5168a7a
# Tricks for Training
## Choice of Optimizers:
Since the development of deep learning, there have been many researchers working on the optimizer. The purpose of the optimizer is to make the loss function as small as possible, so as to find suitable parameters to complete a certain task. At present, the main optimizers used in model training are SGD, RMSProp, Adam, AdaDelt and so on. The SGD optimizers with momentum is widely used in academia and industry, so most of models we release are trained by SGD optimizer with momentum. But the SGD optimizer with momentum has two disadvantages, one is that the convergence speed is slow, the other is that the initial learning rate is difficult to set, however, if the initial learning rate is set properly and the models are trained in sufficient iterations, the models trained by SGD with momentum can reach higher accuracy compared with the models trained by other optimizers. Some other optimizers with adaptive learning rate such as Adam, RMSProp and so on tent to converge faster, but the final convergence accuracy will be slightly worse. If you want to train a model in faster convergence speed, we recommend you use the optimizers with adaptive learning rate, but if you want to train a model with higher accuracy, we recommend you to use SGD optimizer with momentum.
## Choice of Learning Rate and Learning Rate Declining Strategy:
The choice of learning rate is related to the optimizer, data set and tasks. Here we mainly introduce the learning rate of training ImageNet-1K with momentum + SGD as the optimizer and the choice of learning rate decline.
### Concept of Learning Rate：
the learning rate is the hyperparameter to control the learning speed, the lower the learning rate, the slower the change of the loss value, though using a low learning rate can ensure that you will not miss any local minimum, but it also means that the convergence speed is slow, especially when the gradient is trapped in a gradient plateau area.
### Learning Rate Decline Strategy：
During training, if we always use the same learning rate, we cannot get the model with highest accuracy, so the learning rate should be adjust during training. In the early stage of training, the weights are in a random initialization state and the gradients are tended to descent, so we can set a relatively large learning rate for faster convergence. In the late stage of training, the weights are close to the optimal values, the optimal value cannot be reached by a relatively large learning rate, so a relatively smaller learning rate should be used. During training, many researchers use the piecewise_decay learning rate reduction strategy, which is a stepwise decline learning rate. For example, in the training of ResNet50, the initial learning rate we set is 0.1, and the learning rate drops to 1/10 every 30 epoches, the total epoches for training is 120. Besides the piecewise_decay, many researchers also proposed other ways to decrease the learning rate, such as polynomial_decay, exponential_decay and cosine_decay and so on, among them, cosine_decay has become the preferred learning rate reduction method for improving model accuracy beacause there is no need to adjust hyperparameters and the robustness is relatively high. The learning rate curves of cosine_decay and piecewise_decay are shown in the following figures, it is easy to observe that during the entire training process, cosine_decay keeps a relatively large learning rate, so its convergence is slower, but the final convergence accuracy is better than the one using piecewise_decay.
![](../../images/models/lr_decay.jpeg)
In addition, we can also see from the figures that the number of epoches with a small learning rate in cosine_decay is fewer, which will affect the final accuracy, so in order to make cosine_decay play a better effect, it is recommended to use cosine_decay in large epoched, such as 200 epoches.
### Warmup Strategy
If a large batch_size is adopted to train nerual network, we recommend you to adopt warmup strategy. as the name suggests, the warmup strategy is to let model learning first warm up, we do not directly use the initial learning rate at the begining of training, instead, we use a gradually increasing learning rate to train the model, when the increasing learning rate reaches the initial learning rate, the learning rate reduction method mentioned in the learning rate reduction strategy is then used to decay the learning rate. Experiments show that when the batch size is large, warmup strategy can improve the accuracy. Some model training with large batch_size such as MobileNetV3 training, we set the epoch in warmup to 5 by default, that is, first in 5 epoches, the learning rate increases from 0 to initial learning rate, then learning rate decay begins.
## Choice of Batch_size
Batch_size is an important hyperparameter in training neural networks, batch_size determines how much data is sent to the neural network to for training at a time. In the paper [1], the author found in experiments that when batch_size is linearly related to the learning rate, the convergence accuracy is hardly affected. When training ImageNet data, an initial learning rate of 0.1 are commonly chosen for training, and batch_size is 256, so according to the actual model size and memory, you can set the learning rate to 0.1\*k, batch_size to 256\*k.
## Choice of Weight_decay
Overfitting is a common term in machine learning. A simple understanding is that the model performs well on the training data, but it performs poorly on the test data. In the convolutional neural network, there also exists the problem of overfitting. To avoid overfitting, many regular ways have been proposed. Among them, weight_decay is one of the widely used ways to avoid overfitting. After the final loss function, L2 regularization(weight_decay) is added to the loss function, with the help of L2 regularization, the weight of the network tend to choose a smaller value, and finally the parameters in the entire network tends to 0, and the generalization performance of the model is improved accordingly. In different kinds of Deep learning frame, the meaning of L2_decay is the coefficient of L2 regularization, on paddle, the name of this value is L2_decay, so in the following the value is called L2_decay. the larger the coefficient, the more the model tends to be underfitting. In the task of training ImageNet, this parameter is set to 1e-4 in most network. In some small networks such as MobileNet networks, in order to avoid network underfitting, the value is set to 1e-5 ~ 4e-5. Of course, the setting of this value is also related to the specific data set, When the data set is large, the network itself tends to be under-fitted, and the value can be appropriately reduced. When the data set is small, the network tends to overfit itself, so the value can be increased appropriately. The following table shows the accuracy of MobileNetV1_x0_25 using different l2_decay on ImageNet-1k. Since MobileNetV1_x0_25 is a relatively small network, the large l2_decay will make the network tend to be underfitting, so in this network, 3e-5 are better choices compared with 1e-4.
| Model | L2_decay | Train acc1/acc5 | Test acc1/acc5 |
|:--:|:--:|:--:|:--:|
| MobileNetV1_x0_25 | 1e-4 | 43.79%/67.61% | 50.41%/74.70% |
| MobileNetV1_x0_25 | 3e-5 | 47.38%/70.83% | 51.45%/75.45% |
In addition, the setting of L2_decay is also related to whether other regularization is used during training. If the data argument during the training is more complicated, which means that the training becomes more difficult, L2_decay can be appropriately reduced. The following table shows that the precision of ResNet50 using a different l2_decay on ImageNet-1K. It is easy to observe that after the training becomes difficult, using a smaller l2_decay helps to improve the accuracy of the model.
| Model | L2_decay | Train acc1/acc5 | Test acc1/acc5 |
|:--:|:--:|:--:|:--:|
| ResNet50 | 1e-4 | 75.13%/90.42% | 77.65%/93.79% |
| ResNet50 | 7e-5 | 75.56%/90.55% | 78.04%/93.74% |
In summary, l2_decay can be adjusted according to specific tasks and models. Usually simple tasks or larger models are recommended to use Larger l2_decay, complex tasks or smaller models are recommended to use smaller l2_decay.
## Choice of Label_smoothing
Label_smoothing is a regularization method in deep learning. Its full name is Label Smoothing Regularization (LSR), that is, label smoothing regularization. In the traditional classification task, when calculating the loss function, the real one hot label and the output of the neural network are calculated in cross-entropy formula, the label smoothing aims to make the real one hot label become smooth label, which makes the neural network no longer learn from the hard labels, but the soft labels with a probability value, where the probability of the position corresponding to the category is the largest and the probability of other positions are very small value, specific calculation method can be seen in the paper[2]. In label-smoothing, there is an epsilon parameter describing the degree of softening the label. The larger epsilon, the smaller the probability and smoother the label, on the contrary, the label tends to be hard label. during training on ImageNet-1K, the parameter is usually set to 0.1. In the experiments of training ResNet50, when using label_smoothing, the accuracy is higher than the one without label_smoothing, the following table shows the performance of ResNet50_vd with label smoothing and without label smoothing.
| Model | Use_label_smoothing | Test acc1 |
|:--:|:--:|:--:|
| ResNet50_vd | 0 | 77.9% |
| ResNet50_vd | 1 | 78.4% |
But, because label smoothing can be regarded as a regular way, on relatively small models, the accuracy improvement is not obvious or even decreases, the following table shows the accuracy performance of ResNet18 with label smoothing and without label smoothing on ImageNet-1K, it can be clearly seen that after using label smoothing, the accuracy of ResNet has decreased.
| Model | Use_label_smoohing | Train acc1/acc5 | Test acc1/acc5 |
|:--:|:--:|:--:|:--:|
| ResNet18 | 0 | 69.81%/87.70% | 70.98%/89.92% |
| ResNet18 | 1 | 68.00%/86.56% | 70.81%/89.89% |
In summary, the use of label_smoohing for larger models can effectively improve the accuracy of the model, and the use of label_smoohing for smaller models may reduce the accuracy of the model, so before deciding whether to use label_smoohing, you need to evaluate the size of the model and the difficulty of the task.
## Change the Crop Area and Stretch Transformation Degree of the Images for Small Models
In the standard preprocessing of ImageNet-1k data, two values of scale and ratio are defined in the random_crop function. These two values respectively determine the size of the image crop and the degree of stretching of the image. The default value of scale is 0.08-1(lower_scale-upper_scale), the default value range of ratio is 3/4-4/3(lower_ratio-upper_ratio). In small network training, such data argument will make the network underfitting, resulting in a decrease in accuracy. In order to improve the accuracy of the network, you can make the data argument weaker, that is, increase the crop area of the images or weaken the degree of stretching and transformation of the images, we can achieve weaker image transformation by increasing the value of lower_scale or narrowing the gap between lower_ratio and upper_scale. The following table lists the accuracy of training MobileNetV2_x0_25 with different lower_scale. It can be seen that the training accuracy and validation accuracy are improved after increasing the crop area of the images
| Model | Scale Range | Train_acc1/acc5 | Test_acc1/acc5 |
|:--:|:--:|:--:|:--:|
| MobileNetV2_x0_25 | [0.08,1] | 50.36%/72.98% | 52.35%/75.65% |
| MobileNetV2_x0_25 | [0.2,1] | 54.39%/77.08% | 53.18%/76.14% |
## Use Data Augmentation to Improve Accuracy
In general, the size of the data set is critical to the performances, but the annotation of images are often more expensive, so the number of annotated images are often scarce. In this case, the data argument is particularly important. In the standard data augmentation for training on ImageNet-1k, two data augmentation methods which are random_crop and random_flip are mainly used. However, in recent years, more and more data augmentation methods have been proposed, such as cutout, mixup, cutmix, AutoAugment, etc. Experiments show that these data augmentation methods can effectively improve the accuracy of the model. The following table lists the performance of ResNet50 in 8 different data augmentation methods. It can be seen that compared to the baseline, all data augmentation methods can be useful for the accuracy of ResNet50, among them cutmix is currently the most effective data argument. More data argument can be seen here[**Data Argument**](https://paddleclas.readthedocs.io/zh_CN/latest/advanced_tutorials/image_augmentation/ImageAugment.html).
| Model | Data Argument | Test top-1 |
|:--:|:--:|:--:|
| ResNet50 | Baseline | 77.31% |
| ResNet50 | Auto-Augment | 77.95% |
| ResNet50 | Mixup | 78.28% |
| ResNet50 | Cutmix | 78.39% |
| ResNet50 | Cutout | 78.01% |
| ResNet50 | Gridmask | 77.85% |
| ResNet50 | Random-Augment | 77.70% |
| ResNet50 | Random-Erasing | 77.91% |
| ResNet50 | Hide-and-Seek | 77.43% |
## Determine the Tuning Strategy by Train_acc and Test_acc
In the process of training the network, the training set accuracy rate and validation set accuracy rate of each epoch are usually printed. Generally speaking, the accuracy of the training set is slightly higher than the accuracy of the validation set or the same are good state in training, but if you find that the accuracy of training set is much higher than the one of validation set, it means that overfitting happens in your task, which need more regularization, such as increase the value of L2_decay, using more data argument or label smoothing and so on. If you find that the accuracy of training set is lower than the one of validation set, it means that underfitting happens in your task, which recommend you to decrease the value of L2_decay, using fewer data argument, increase the area of the crop area of the images, weaken the stretching transformation of the images, remove label_smoothing, etc.
## Improve the Accuracy of Your Own Data Set with Existing Pre-trained Models
In the field of computer vision, it has become common to load pre-trained models to train one's own tasks. Compared with starting training from random initialization, loading pre-trained models can often improve the accuracy of specific tasks. In general, the pre-trained model widely used in the industry is obtained from the ImageNet-1k dataset. The fc layer weight of the pre-trained model is a matrix of k\*1000, where k is The number of neurons before, and the weights of the fc layer is not need to load because of the different tasks. In terms of learning rate, if your training data set is particularly small (such as less than 1,000), we recommend that you use a smaller initial learning rate, such as 0.001 (batch_size: 256, the same below), to avoid a large learning rate undermine pre-training weights, if your training data set is relatively large (greater than 100,000), we recommend that you try a larger initial learning rate, such as 0.01 or greater.
> If you think this guide is helpful to you, welcome to star our repo:[https://github.com/PaddlePaddle/PaddleClas](https://github.com/PaddlePaddle/PaddleClas)
## Reference
[1]P. Goyal, P. Dolla ́r, R. B. Girshick, P. Noordhuis, L. Wesolowski, A. Kyrola, A. Tulloch, Y. Jia, and K. He. Accurate, large minibatch SGD: training imagenet in 1 hour. CoRR, abs/1706.02677, 2017.
[2]C.Szegedy,V.Vanhoucke,S.Ioffe,J.Shlens,andZ.Wojna. Rethinking the inception architecture for computer vision. CoRR, abs/1512.00567, 2015.
docs/en/models/index.rst
浏览文件 @ f5168a7a
......@@ -4,13 +4,13 @@ models
.. toctree::
:maxdepth: 1
models_intro.md
Tricks.md
ResNet_and_vd.md
Mobile.md
SEResNext_and_Res2Net.md
Inception.md
HRNet.md
DPN_DenseNet.md
EfficientNet_and_ResNeXt101_wsl.md
Others.md
models_intro_en.md
Tricks_en.md
ResNet_and_vd_en.md
Mobile_en.md
SEResNext_and_Res2Net_en.md
Inception_en.md
HRNet_en.md
DPN_DenseNet_en.md
EfficientNet_and_ResNeXt101_wsl_en.md
Others_en.md
docs/en/models/models_intro_en.md 0 → 100644
浏览文件 @ f5168a7a
此差异已折叠。
docs/en/tutorials/config.md → docs/en/tutorials/config_en.md
浏览文件 @ f5168a7a
......@@ -4,9 +4,7 @@
## Introduction
This document introduces the configuration(filed in config/*.yaml) of PaddleClas.
## Filed in config/*.yaml
This document introduces the configuration(filed in `config/*.yaml`) of PaddleClas.
### Basic
......
docs/en/tutorials/data.md → docs/en/tutorials/data_en.md
浏览文件 @ f5168a7a
......@@ -2,10 +2,10 @@
---
## 1. Introducation
## Introducation
This document introduces the preparation of ImageNet1k and flowers102
## 2. Dataset
## Dataset
Dataset | train dataset size | valid dataset size | category |
:------:|:---------------:|:---------------------:|:--------:|
......
docs/en/tutorials/getting_started.md → docs/en/tutorials/getting_started_en.md
浏览文件 @ f5168a7a
......@@ -64,7 +64,7 @@ visualdl --logdir ./scalar --host <host_IP> --port <port_num>
* please refer to [Trial](./quick_start.md) for more details.
### validating
### validation
```bash
python tools/eval.py \
......
docs/en/tutorials/index.rst
浏览文件 @ f5168a7a
......@@ -4,8 +4,8 @@ tutorials
.. toctree::
:maxdepth: 1
install.md
quick_start.md
data.md
getting_started.md
config.md
install_en.md
quick_start_en.md
data_en.md
getting_started_en.md
config_en.md
docs/en/update_history_en.md 0 → 100644
浏览文件 @ f5168a7a
# Release Notes
* 2020.06.17
* Add English documents。
* 2020.06.12
* Add support for training and evaluation on Windows or CPU.
* 2020.05.17
* Add support for mixed precision training.
* 2020.05.09
* Add user guide about Paddle Serving and Paddle-Lite.
* Add benchmark about FP16/FP32 on T4 GPU.
* 2020.04.14
* First commit.
docs/images/main_features_s.png
查看替换文件 @ f03fab0e 浏览文件 @ f5168a7a
docs/images/main_features_s.png

154.4 KB | W: | H:

docs/images/main_features_s.png

275.4 KB | W: | H:

docs/images/main_features_s.png
docs/images/main_features_s.png
docs/images/main_features_s.png
docs/images/main_features_s.png
  • 2-up
  • Swipe
  • Onion skin
docs/images/main_features_s_en.png
查看替换文件 @ f03fab0e 浏览文件 @ f5168a7a
docs/images/main_features_s_en.png

103.1 KB | W: | H:

docs/images/main_features_s_en.png

200.0 KB | W: | H:

docs/images/main_features_s_en.png
docs/images/main_features_s_en.png
docs/images/main_features_s_en.png
docs/images/main_features_s_en.png
  • 2-up
  • Swipe
  • Onion skin
docs/zh_CN/advanced_tutorials/distillation/distillation.md
浏览文件 @ f5168a7a
......@@ -8,7 +8,7 @@
深度神经网络一般有较多的参数冗余，目前有几种主要的方法对模型进行压缩，减小其参数量。如裁剪、量化、知识蒸馏等，其中知识蒸馏是指使用教师模型(teacher model)去指导学生模型(student model)学习特定任务，保证小模型在参数量不变的情况下，得到比较大的性能提升，甚至获得与大模型相似的精度指标[1]。PaddleClas融合已有的蒸馏方法[2,3]，提供了一种简单的半监督标签知识蒸馏方案（SSLD，Simple Semi-supervised Label Distillation），基于ImageNet1k分类数据集，在ResNet_vd以及MobileNet系列上的精度均有超过3%的绝对精度提升，具体指标如下图所示。
![](../../../images/distillation/distillation_perform.png)
![](../../../images/distillation/distillation_perform_s.jpg)
# 二、SSLD 蒸馏策略
......@@ -17,10 +17,8 @@
SSLD的流程图如下图所示。
![](../../../images/distillation/ppcls_distillation.png)
首先，我们从ImageNet22k中挖掘出了近400万张图片，同时与ImageNet-1k训练集整合在一起，得到了一个新的包含500万张图片的数据集。然后，我们将学生模型与教师模型组合成一个新的网络，该网络分别输出学生模型和教师模型的预测分布，与此同时，固定教师模型整个网络的梯度，而学生模型可以做正常的反向传播。最后，我们将两个模型的logits经过softmax激活函数转换为soft label，并将二者的soft label做JS散度作为损失函数，用于蒸馏模型训练。下面以MobileNetV3（该模型直接训练，精度为75.3%）的知识蒸馏为例，介绍该方案的核心关键点（baseline为79.12%的ResNet50_vd模型蒸馏MobileNetV3，训练集为ImageNet1k训练集，loss为cross entropy loss，迭代轮数为120epoch，精度指标为75.6%）。
* 教师模型的选择。在进行知识蒸馏时，如果教师模型与学生模型的结构差异太大，蒸馏得到的结果反而不会有太大收益。相同结构下，精度更高的教师模型对结果也有很大影响。相比于79.12%的ResNet50_vd教师模型，使用82.4%的ResNet50_vd教师模型可以带来0.4%的绝对精度收益(`75.6%->76.0%`)。
......@@ -103,6 +101,14 @@ SSLD的流程图如下图所示。
| ResNet101_vd | 30 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 83.73% |
## 3.4 数据增广以及基于Fix策略的微调
* 基于前文所述的实验结论，我们在训练的过程中加入自动增广(AutoAugment)[4]，同时进一步减小了l2_decay(4e-5->2e-5)，最终ResNet50_vd经过SSLD蒸馏策略，在ImageNet1k上的精度可以达到82.99%，相比之前不加数据增广的蒸馏策略再次增加了0.6%。
* 对于图像分类任务，在测试的时候，测试尺度为训练尺度的1.15倍左右时，往往在不需要重新训练模型的情况下，模型的精度指标就可以进一步提升[5]，对于82.99%的ResNet50_vd在320x320的尺度下测试，精度可达83.7%，我们进一步使用Fix策略，即在320x320的尺度下进行训练，使用与预测时相同的数据预处理方法，同时固定除FC层以外的所有参数，最终在320x320的预测尺度下，精度可以达到**84.0%**。
## 3.4 实验过程中的一些问题
### 3.4.1 bn的计算方法
......@@ -182,7 +188,7 @@ for var in ./*_student; do cp "$var" "../student_model/${var%_student}"; done #
| Faster RCNN R50_vd FPN | 640/640 | 79.12% | [0.05,0.05,0.1,0.1,0.15] | 34.3% |
| Faster RCNN R50_vd FPN | 640/640 | 82.18% | [0.05,0.05,0.1,0.1,0.15] | 36.3% |
在这里可以看出，对于未蒸馏模型，过度调整中间层学习率反而降低最终检测模型的性能指标。基于该蒸馏模型，我们也提供了领先的服务端实用目标检测方案，详细的配置与训练代码均已开源，可以参考[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/rcnn_server_side_det)。
在这里可以看出，对于未蒸馏模型，过度调整中间层学习率反而降低最终检测模型的性能指标。基于该蒸馏模型，我们也提供了领先的服务端实用目标检测方案，详细的配置与训练代码均已开源，可以参考[PaddleDetection](https://github.com/PaddlePaddle/PaddleDetection/tree/master/configs/rcnn_enhance)。
# 五、SSLD实战
......@@ -266,3 +272,7 @@ sh tools/run.sh
[2] Bagherinezhad H, Horton M, Rastegari M, et al. Label refinery: Improving imagenet classification through label progression[J]. arXiv preprint arXiv:1805.02641, 2018.
[3] Yalniz I Z, Jégou H, Chen K, et al. Billion-scale semi-supervised learning for image classification[J]. arXiv preprint arXiv:1905.00546, 2019.
[4] Cubuk E D, Zoph B, Mane D, et al. Autoaugment: Learning augmentation strategies from data[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2019: 113-123.
[5] Touvron H, Vedaldi A, Douze M, et al. Fixing the train-test resolution discrepancy[C]//Advances in Neural Information Processing Systems. 2019: 8250-8260.
docs/zh_CN/update_history.md
浏览文件 @ f5168a7a
# 更新日志
* 2020.06.17
* 添加英文文档。
* 2020.06.12
* 添加对windows和CPU环境的训练与评估支持。
* 2020.05.17
* 添加混合精度训练。
......
ppcls/data/reader.py
浏览文件 @ f5168a7a
......@@ -15,15 +15,17 @@
import numpy as np
import imghdr
import os
import sys
import signal
from paddle import fluid
from paddle.fluid.io import multiprocess_reader
from . import imaug
from .imaug import transform
from ppcls.utils import logger
trainers_num = int(os.environ.get('PADDLE_TRAINERS_NUM', 1))
trainers_num = int(os.environ.get('PADDLE_TRAINERS_NUM', 0))
trainer_id = int(os.environ.get("PADDLE_TRAINER_ID", 0))
......@@ -139,8 +141,9 @@ def get_file_list(params):
# use only partial data for each trainer in distributed training
if params['mode'] == 'train':
img_per_trainer = len(full_lines) // trainers_num
full_lines = full_lines[trainer_id::trainers_num][:img_per_trainer]
real_trainer_num = max(trainers_num, 1)
img_per_trainer = len(full_lines) // real_trainer_num
full_lines = full_lines[trainer_id::real_trainer_num][:img_per_trainer]
return full_lines
......@@ -165,7 +168,7 @@ def create_operators(params):
return ops
def partial_reader(params, full_lines, part_id=0, part_num=1):
def partial_reader(params, full_lines, part_id=0, part_num=1, batch_size=1):
"""
create a reader with partial data
......@@ -174,13 +177,13 @@ def partial_reader(params, full_lines, part_id=0, part_num=1):
full_lines: label list
part_id(int): part index of the current partial data
part_num(int): part num of the dataset
batch_size(int): batch size for one trainer
"""
assert part_id < part_num, ("part_num: {} should be larger "
"than part_id: {}".format(part_num, part_id))
full_lines = full_lines[part_id::part_num]
batch_size = int(params['batch_size']) // trainers_num
if params['mode'] != "test" and len(full_lines) < batch_size:
raise SampleNumException('', len(full_lines), batch_size)
......@@ -197,7 +200,7 @@ def partial_reader(params, full_lines, part_id=0, part_num=1):
return reader
def mp_reader(params):
def mp_reader(params, batch_size):
"""
multiprocess reader
......@@ -210,11 +213,16 @@ def mp_reader(params):
if params["mode"] == "train":
full_lines = shuffle_lines(full_lines, seed=None)
# NOTE: multiprocess reader is not supported on windows
if sys.platform == "win32":
return partial_reader(params, full_lines, 0, 1, batch_size)
part_num = 1 if 'num_workers' not in params else params['num_workers']
readers = []
for part_id in range(part_num):
readers.append(partial_reader(params, full_lines, part_id, part_num))
readers.append(
partial_reader(params, full_lines, part_id, part_num, batch_size))
return multiprocess_reader(readers, use_pipe=False)
......@@ -248,6 +256,7 @@ class Reader:
except KeyError:
raise ModeException(mode=mode)
self.use_gpu = config.get("use_gpu", True)
use_mix = config.get('use_mix')
self.params['mode'] = mode
if seed is not None:
......@@ -257,10 +266,17 @@ class Reader:
self.batch_ops = create_operators(self.params['mix'])
def __call__(self):
batch_size = int(self.params['batch_size']) // trainers_num
device_num = trainers_num
# non-distributed launch
if trainers_num <= 0:
if self.use_gpu:
device_num = fluid.core.get_cuda_device_count()
else:
device_num = int(os.environ.get('CPU_NUM', 1))
batch_size = int(self.params['batch_size']) // device_num
def wrapper():
reader = mp_reader(self.params)
reader = mp_reader(self.params, batch_size)
batch = []
for idx, sample in enumerate(reader()):
img, label = sample
......
ppcls/modeling/architectures/__init__.py
浏览文件 @ f5168a7a
......@@ -42,8 +42,9 @@ from .res2net_vd import Res2Net50_vd_48w_2s, Res2Net50_vd_26w_4s, Res2Net50_vd_1
from .hrnet import HRNet_W18_C, HRNet_W30_C, HRNet_W32_C, HRNet_W40_C, HRNet_W44_C, HRNet_W48_C, HRNet_W60_C, HRNet_W64_C, SE_HRNet_W18_C, SE_HRNet_W30_C, SE_HRNet_W32_C, SE_HRNet_W40_C, SE_HRNet_W44_C, SE_HRNet_W48_C, SE_HRNet_W60_C, SE_HRNet_W64_C
from .darts_gs import DARTS_GS_6M, DARTS_GS_4M
from .resnet_acnet import ResNet18_ACNet, ResNet34_ACNet, ResNet50_ACNet, ResNet101_ACNet, ResNet152_ACNet
from .ghostnet import GhostNet_x0_5, GhostNet_x1_0, GhostNet_x1_3
# distillation model
from .distillation_models import ResNet50_vd_distill_MobileNetV3_large_x1_0, ResNeXt101_32x16d_wsl_distill_ResNet50_vd
from .csp_resnet import CSPResNet50_leaky
\ No newline at end of file
from .csp_resnet import CSPResNet50_leaky
ppcls/modeling/architectures/ghostnet.py 0 → 100644
浏览文件 @ f5168a7a
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import math
import paddle.fluid as fluid
from paddle.fluid.param_attr import ParamAttr
__all__ = ["GhostNet", "GhostNet_x0_5", "GhostNet_x1_0", "GhostNet_x1_3"]
class GhostNet():
def __init__(self, scale):
cfgs = [
# k, t, c, SE, s
[3, 16, 16, 0, 1],
[3, 48, 24, 0, 2],
[3, 72, 24, 0, 1],
[5, 72, 40, 1, 2],
[5, 120, 40, 1, 1],
[3, 240, 80, 0, 2],
[3, 200, 80, 0, 1],
[3, 184, 80, 0, 1],
[3, 184, 80, 0, 1],
[3, 480, 112, 1, 1],
[3, 672, 112, 1, 1],
[5, 672, 160, 1, 2],
[5, 960, 160, 0, 1],
[5, 960, 160, 1, 1],
[5, 960, 160, 0, 1],
[5, 960, 160, 1, 1]
]
self.cfgs = cfgs
self.scale = scale
def net(self, input, class_dim=1000):
# build first layer:
output_channel = int(self._make_divisible(16 * self.scale, 4))
x = self.conv_bn_layer(input=input,
num_filters=output_channel,
filter_size=3,
stride=2,
groups=1,
act="relu",
name="conv1")
# build inverted residual blocks
idx = 0
for k, exp_size, c, use_se, s in self.cfgs:
output_channel = int(self._make_divisible(c * self.scale, 4))
hidden_channel = int(self._make_divisible(exp_size * self.scale, 4))
x = self.ghost_bottleneck(input=x,
hidden_dim=hidden_channel,
output=output_channel,
kernel_size=k,
stride=s,
use_se=use_se,
name="_ghostbottleneck_" + str(idx))
idx += 1
# build last several layers
output_channel = int(self._make_divisible(exp_size * self.scale, 4))
x = self.conv_bn_layer(input=x,
num_filters=output_channel,
filter_size=1,
stride=1,
groups=1,
act="relu",
name="conv_last")
x = fluid.layers.pool2d(input=x, pool_type='avg', global_pooling=True)
output_channel = 1280
stdv = 1.0 / math.sqrt(x.shape[1] * 1.0)
out = self.conv_bn_layer(input=x,
num_filters=output_channel,
filter_size=1,
stride=1,
act="relu",
name="fc_0")
out = fluid.layers.dropout(x=out, dropout_prob=0.2)
stdv = 1.0 / math.sqrt(out.shape[1] * 1.0)
out = fluid.layers.fc(input=out,
size=class_dim,
param_attr=ParamAttr(name="fc_1_weights",
initializer=fluid.initializer.Uniform(-stdv, stdv)),
bias_attr=ParamAttr(name="fc_1_offset"))
return out
def _make_divisible(self, v, divisor, min_value=None):
"""
This function is taken from the original tf repo.
It ensures that all layers have a channel number that is divisible by 8
It can be seen here:
https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py
"""
if min_value is None:
min_value = divisor
new_v = max(min_value, int(v + divisor / 2) // divisor * divisor)
# Make sure that round down does not go down by more than 10%.
if new_v < 0.9 * v:
new_v += divisor
return new_v
def conv_bn_layer(self,
input,
num_filters,
filter_size,
stride=1,
groups=1,
act=None,
name=None):
x = fluid.layers.conv2d(input=input,
num_filters=num_filters,
filter_size=filter_size,
stride=stride,
padding=(filter_size - 1) // 2,
groups=groups,
act=None,
param_attr=ParamAttr(
initializer=fluid.initializer.MSRA(), name=name + "_weights"),
bias_attr=False)
bn_name = name + "_bn"
x = fluid.layers.batch_norm(input=x,
act=act,
param_attr=ParamAttr(
name=bn_name + "_scale",
regularizer=fluid.regularizer.L2DecayRegularizer(
regularization_coeff=0.0)),
bias_attr=ParamAttr(
name=bn_name + "_offset",
regularizer=fluid.regularizer.L2DecayRegularizer(
regularization_coeff=0.0)),
moving_mean_name=bn_name + "_mean",
moving_variance_name=name + "_variance")
return x
def se_block(self, input, num_channels, reduction_ratio=4, name=None):
pool = fluid.layers.pool2d(input=input, pool_type='avg', global_pooling=True, use_cudnn=False)
stdv = 1.0 / math.sqrt(pool.shape[1] * 1.0)
squeeze = fluid.layers.fc(input=pool,
size=num_channels // reduction_ratio,
act='relu',
param_attr=fluid.param_attr.ParamAttr(
initializer=fluid.initializer.Uniform(-stdv, stdv),
name=name + '_1_weights'),
bias_attr=ParamAttr(name=name + '_1_offset'))
stdv = 1.0 / math.sqrt(squeeze.shape[1] * 1.0)
excitation = fluid.layers.fc(input=squeeze,
size=num_channels,
act=None,
param_attr=fluid.param_attr.ParamAttr(
initializer=fluid.initializer.Uniform(-stdv, stdv),
name=name + '_2_weights'),
bias_attr=ParamAttr(name=name + '_2_offset'))
#excitation = fluid.layers.clip(x=excitation, min=0, max=1)
se_scale = fluid.layers.elementwise_mul(x=input, y=excitation, axis=0)
return se_scale
def depthwise_conv(self,
input,
output,
kernel_size,
stride=1,
relu=False,
name=None):
return self.conv_bn_layer(input=input,
num_filters=output,
filter_size=kernel_size,
stride=stride,
groups=input.shape[1],
act="relu" if relu else None,
name=name + "_depthwise")
def ghost_module(self,
input,
output,
kernel_size=1,
ratio=2,
dw_size=3,
stride=1,
relu=True,
name=None):
self.output = output
init_channels = int(math.ceil(output / ratio))
new_channels = int(init_channels * (ratio - 1))
primary_conv = self.conv_bn_layer(input=input,
num_filters=init_channels,
filter_size=kernel_size,
stride=stride,
groups=1,
act="relu" if relu else None,
name=name + "_primary_conv")
cheap_operation = self.conv_bn_layer(input=primary_conv,
num_filters=new_channels,
filter_size=dw_size,
stride=1,
groups=init_channels,
act="relu" if relu else None,
name=name + "_cheap_operation")
out = fluid.layers.concat([primary_conv, cheap_operation], axis=1)
return out
def ghost_bottleneck(self,
input,
hidden_dim,
output,
kernel_size,
stride,
use_se,
name=None):
inp_channels = input.shape[1]
x = self.ghost_module(input=input,
output=hidden_dim,
kernel_size=1,
stride=1,
relu=True,
name=name + "_ghost_module_1")
if stride == 2:
x = self.depthwise_conv(input=x,
output=hidden_dim,
kernel_size=kernel_size,
stride=stride,
relu=False,
name=name + "_depthwise")
if use_se:
x = self.se_block(input=x, num_channels=hidden_dim, name=name + "_se")
x = self.ghost_module(input=x,
output=output,
kernel_size=1,
relu=False,
name=name + "_ghost_module_2")
if stride == 1 and inp_channels == output:
shortcut = input
else:
shortcut = self.depthwise_conv(input=input,
output=inp_channels,
kernel_size=kernel_size,
stride=stride,
relu=False,
name=name + "_shortcut_depthwise")
shortcut = self.conv_bn_layer(input=shortcut,
num_filters=output,
filter_size=1,
stride=1,
groups=1,
act=None,
name=name + "_shortcut_conv")
return fluid.layers.elementwise_add(x=x,
y=shortcut,
axis=-1)
def GhostNet_x0_5():
model = GhostNet(scale=0.5)
return model
def GhostNet_x1_0():
model = GhostNet(scale=1.0)
return model
def GhostNet_x1_3():
model = GhostNet(scale=1.3)
return model
ppcls/modeling/architectures/hrnet.py
浏览文件 @ f5168a7a
#copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
#Licensed under the Apache License, Version 2.0 (the "License");
#you may not use this file except in compliance with the License.
#You may obtain a copy of the License at
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
#Unless required by applicable law or agreed to in writing, software
#distributed under the License is distributed on an "AS IS" BASIS,
#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#See the License for the specific language governing permissions and
#limitations under the License.
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
......@@ -74,7 +74,7 @@ class HRNet():
tr3 = self.transition_layer(st3, channels_3, channels_4, name='tr3')
st4 = self.stage(tr3, num_modules_4, channels_4, name='st4')
#classification
# classification
last_cls = self.last_cls_out(x=st4, name='cls_head')
y = last_cls[0]
last_num_filters = [256, 512, 1024]
......@@ -273,7 +273,7 @@ class HRNet():
input=conv,
num_channels=num_filters,
reduction_ratio=16,
name=name + '_fc')
name="fc" + name)
return fluid.layers.elementwise_add(x=residual, y=conv, act='relu')
def bottleneck_block(self,
......@@ -312,7 +312,7 @@ class HRNet():
input=conv,
num_channels=num_filters * 4,
reduction_ratio=16,
name=name + '_fc')
name="fc" + name)
return fluid.layers.elementwise_add(x=residual, y=conv, act='relu')
def squeeze_excitation(self,
......@@ -325,7 +325,7 @@ class HRNet():
stdv = 1.0 / math.sqrt(pool.shape[1] * 1.0)
squeeze = fluid.layers.fc(
input=pool,
size=num_channels / reduction_ratio,
size=int(num_channels / reduction_ratio),
act='relu',
param_attr=fluid.param_attr.ParamAttr(
initializer=fluid.initializer.Uniform(-stdv, stdv),
......
ppcls/modeling/architectures/resnet_vd.py
浏览文件 @ f5168a7a
#copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
#Licensed under the Apache License, Version 2.0 (the "License");
#you may not use this file except in compliance with the License.
#You may obtain a copy of the License at
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
#Unless required by applicable law or agreed to in writing, software
#distributed under the License is distributed on an "AS IS" BASIS,
#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#See the License for the specific language governing permissions and
#limitations under the License.
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
......@@ -49,7 +49,8 @@ class ResNet():
layers = self.layers
supported_layers = [18, 34, 50, 101, 152, 200]
assert layers in supported_layers, \
"supported layers are {} but input layer is {}".format(supported_layers, layers)
"supported layers are {} but input layer is {}".format(
supported_layers, layers)
if layers == 18:
depth = [2, 2, 2, 2]
......@@ -159,7 +160,9 @@ class ResNet():
padding=(filter_size - 1) // 2,
groups=groups,
act=None,
param_attr=ParamAttr(name=name + "_weights" + self.postfix_name),
param_attr=ParamAttr(
name=name + "_weights" + self.postfix_name,
learning_rate=lr_mult),
bias_attr=False)
if name == "conv1":
bn_name = "bn_" + name
......@@ -168,8 +171,12 @@ class ResNet():
return fluid.layers.batch_norm(
input=conv,
act=act,
param_attr=ParamAttr(name=bn_name + '_scale' + self.postfix_name),
bias_attr=ParamAttr(bn_name + '_offset' + self.postfix_name),
param_attr=ParamAttr(
name=bn_name + '_scale' + self.postfix_name,
learning_rate=lr_mult),
bias_attr=ParamAttr(
bn_name + '_offset' + self.postfix_name,
learning_rate=lr_mult),
moving_mean_name=bn_name + '_mean' + self.postfix_name,
moving_variance_name=bn_name + '_variance' + self.postfix_name)
......
ppcls/utils/config.py
浏览文件 @ f5168a7a
......@@ -64,14 +64,18 @@ def print_dict(d, delimiter=0):
placeholder = "-" * 60
for k, v in sorted(d.items()):
if isinstance(v, dict):
logger.info("{}{} : ".format(delimiter * " ", logger.coloring(k, "HEADER")))
logger.info("{}{} : ".format(delimiter * " ",
logger.coloring(k, "HEADER")))
print_dict(v, delimiter + 4)
elif isinstance(v, list) and len(v) >= 1 and isinstance(v[0], dict):
logger.info("{}{} : ".format(delimiter * " ", logger.coloring(str(k),"HEADER")))
logger.info("{}{} : ".format(delimiter * " ",
logger.coloring(str(k), "HEADER")))
for value in v:
print_dict(value, delimiter + 4)
else:
logger.info("{}{} : {}".format(delimiter * " ", logger.coloring(k,"HEADER"), logger.coloring(v,"OKGREEN")))
logger.info("{}{} : {}".format(delimiter * " ",
logger.coloring(k, "HEADER"),
logger.coloring(v, "OKGREEN")))
if k.isupper():
logger.info(placeholder)
......@@ -95,7 +99,9 @@ def check_config(config):
check.check_version()
mode = config.get('mode', 'train')
check.check_gpu()
use_gpu = config.get("use_gpu", True)
if use_gpu:
check.check_gpu()
architecture = config.get('ARCHITECTURE')
check.check_architecture(architecture)
......
tools/eval_multi_platform.py 0 → 100644
浏览文件 @ f5168a7a
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import os
import argparse
import paddle.fluid as fluid
import program
from ppcls.data import Reader
from ppcls.utils.config import get_config
from ppcls.utils.save_load import init_model
def parse_args():
parser = argparse.ArgumentParser("PaddleClas eval script")
parser.add_argument(
'-c',
'--config',
type=str,
default='./configs/eval.yaml',
help='config file path')
parser.add_argument(
'-o',
'--override',
action='append',
default=[],
help='config options to be overridden')
args = parser.parse_args()
return args
def main(args):
config = get_config(args.config, overrides=args.override, show=True)
use_gpu = config.get("use_gpu", True)
places = fluid.cuda_places() if use_gpu else fluid.cpu_places()
startup_prog = fluid.Program()
valid_prog = fluid.Program()
valid_dataloader, valid_fetchs = program.build(
config, valid_prog, startup_prog, is_train=False, is_distributed=False)
valid_prog = valid_prog.clone(for_test=True)
exe = fluid.Executor(places[0])
exe.run(startup_prog)
init_model(config, valid_prog, exe)
valid_reader = Reader(config, 'valid')()
valid_dataloader.set_sample_list_generator(valid_reader, places)
compiled_valid_prog = program.compile(config, valid_prog)
program.run(valid_dataloader, exe, compiled_valid_prog, valid_fetchs, -1,
'eval')
if __name__ == '__main__':
args = parse_args()
main(args)
tools/program.py
浏览文件 @ f5168a7a
......@@ -18,6 +18,7 @@ from __future__ import print_function
import os
import time
import numpy as np
from collections import OrderedDict
......@@ -316,7 +317,7 @@ def mixed_precision_optimizer(config, optimizer):
return optimizer
def build(config, main_prog, startup_prog, is_train=True):
def build(config, main_prog, startup_prog, is_train=True, is_distributed=True):
"""
Build a program using a model and an optimizer
1. create feeds
......@@ -330,6 +331,7 @@ def build(config, main_prog, startup_prog, is_train=True):
main_prog(): main program
startup_prog(): startup program
is_train(bool): train or valid
is_distributed(bool): whether to use distributed training method
Returns:
dataloader(): a bridge between the model and the data
......@@ -366,7 +368,8 @@ def build(config, main_prog, startup_prog, is_train=True):
fetchs['lr'] = (lr, AverageMeter('lr', 'f', need_avg=False))
optimizer = mixed_precision_optimizer(config, optimizer)
optimizer = dist_optimizer(config, optimizer)
if is_distributed:
optimizer = dist_optimizer(config, optimizer)
optimizer.minimize(fetchs['loss'][0])
if config.get('use_ema'):
......@@ -380,7 +383,7 @@ def build(config, main_prog, startup_prog, is_train=True):
return dataloader, fetchs
def compile(config, program, loss_name=None):
def compile(config, program, loss_name=None, share_prog=None):
"""
Compile the program
......@@ -388,6 +391,7 @@ def compile(config, program, loss_name=None):
config(dict): config
program(): the program which is wrapped by
loss_name(str): loss name
share_prog(): the shared program, used for evaluation during training
Returns:
compiled_program(): a compiled program
......@@ -399,6 +403,7 @@ def compile(config, program, loss_name=None):
exec_strategy.num_iteration_per_drop_scope = 10
compiled_program = fluid.CompiledProgram(program).with_data_parallel(
share_vars_from=share_prog,
loss_name=loss_name,
build_strategy=build_strategy,
exec_strategy=exec_strategy)
......@@ -443,7 +448,7 @@ def run(dataloader,
batch_time.update(time.time() - tic)
tic = time.time()
for i, m in enumerate(metrics):
metric_list[i].update(m[0], len(batch[0]))
metric_list[i].update(np.mean(m), len(batch[0]))
fetchs_str = ''.join([str(m.value) + ' '
for m in metric_list] + [batch_time.value]) + 's'
if vdl_writer:
......
tools/train.py
浏览文件 @ f5168a7a
......@@ -143,8 +143,7 @@ def main(args):
model_path = os.path.join(config.model_save_dir,
config.ARCHITECTURE["name"])
save_model(train_prog, model_path,
"best_model_in_epoch_" + str(epoch_id))
save_model(train_prog, model_path, "best_model")
# 3. save the persistable model
if epoch_id % config.save_interval == 0:
......
tools/train_multi_platform.py 0 → 100644
浏览文件 @ f5168a7a
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import os
import paddle.fluid as fluid
from ppcls.data import Reader
from ppcls.utils.config import get_config
from ppcls.utils.save_load import init_model, save_model
from ppcls.utils import logger
import program
def parse_args():
parser = argparse.ArgumentParser("PaddleClas train script")
parser.add_argument(
'-c',
'--config',
type=str,
default='configs/ResNet/ResNet50.yaml',
help='config file path')
parser.add_argument(
'--vdl_dir',
type=str,
default=None,
help='VisualDL logging directory for image.')
parser.add_argument(
'-o',
'--override',
action='append',
default=[],
help='config options to be overridden')
args = parser.parse_args()
return args
def main(args):
config = get_config(args.config, overrides=args.override, show=True)
# assign the place
use_gpu = config.get("use_gpu", True)
places = fluid.cuda_places() if use_gpu else fluid.cpu_places()
# startup_prog is used to do some parameter init work,
# and train prog is used to hold the network
startup_prog = fluid.Program()
train_prog = fluid.Program()
best_top1_acc = 0.0 # best top1 acc record
if not config.get('use_ema'):
train_dataloader, train_fetchs = program.build(
config,
train_prog,
startup_prog,
is_train=True,
is_distributed=False)
else:
train_dataloader, train_fetchs, ema = program.build(
config,
train_prog,
startup_prog,
is_train=True,
is_distributed=False)
if config.validate:
valid_prog = fluid.Program()
valid_dataloader, valid_fetchs = program.build(
config,
valid_prog,
startup_prog,
is_train=False,
is_distributed=False)
# clone to prune some content which is irrelevant in valid_prog
valid_prog = valid_prog.clone(for_test=True)
# create the "Executor" with the statement of which place
exe = fluid.Executor(places[0])
# Parameter initialization
exe.run(startup_prog)
# load model from 1. checkpoint to resume training, 2. pretrained model to finetune
init_model(config, train_prog, exe)
train_reader = Reader(config, 'train')()
train_dataloader.set_sample_list_generator(train_reader, places)
compiled_train_prog = program.compile(config, train_prog,
train_fetchs['loss'][0].name)
if config.validate:
valid_reader = Reader(config, 'valid')()
valid_dataloader.set_sample_list_generator(valid_reader, places)
compiled_valid_prog = program.compile(
config, valid_prog, share_prog=compiled_train_prog)
if args.vdl_dir:
from visualdl import LogWriter
vdl_writer = LogWriter(args.vdl_dir)
else:
vdl_writer = None
for epoch_id in range(config.epochs):
# 1. train with train dataset
program.run(train_dataloader, exe, compiled_train_prog, train_fetchs,
epoch_id, 'train', vdl_writer)
# 2. validate with validate dataset
if config.validate and epoch_id % config.valid_interval == 0:
if config.get('use_ema'):
logger.info(logger.coloring("EMA validate start..."))
with ema.apply(exe):
top1_acc = program.run(valid_dataloader, exe,
compiled_valid_prog, valid_fetchs,
epoch_id, 'valid')
logger.info(logger.coloring("EMA validate over!"))
top1_acc = program.run(valid_dataloader, exe, compiled_valid_prog,
valid_fetchs, epoch_id, 'valid')
if top1_acc > best_top1_acc:
best_top1_acc = top1_acc
message = "The best top1 acc {:.5f}, in epoch: {:d}".format(
best_top1_acc, epoch_id)
logger.info("{:s}".format(logger.coloring(message, "RED")))
if epoch_id % config.save_interval == 0:
model_path = os.path.join(config.model_save_dir,
config.ARCHITECTURE["name"])
save_model(train_prog, model_path,
"best_model_in_epoch_" + str(epoch_id))
# 3. save the persistable model
if epoch_id % config.save_interval == 0:
model_path = os.path.join(config.model_save_dir,
config.ARCHITECTURE["name"])
save_model(train_prog, model_path, epoch_id)
if __name__ == '__main__':
args = parse_args()
main(args)
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或