diff --git a/configs/ppyoloe/README.md b/configs/ppyoloe/README.md index 7af1199ae8beeba19aad7415273736b75e061c5c..c7a8b5a9d647c891ec1684bd09001fa668f18ae3 100644 --- a/configs/ppyoloe/README.md +++ b/configs/ppyoloe/README.md @@ -9,7 +9,7 @@ English | [简体中文](README_cn.md) - [Appendix](#Appendix) ## Introduction -PP-YOLOE is an excellent single-stage anchor-free model based on PP-YOLOv2, surpassing a variety of popular yolo models. PP-YOLOE has a series of models, named s/m/l/x, which are configured through width multiplier and depth multiplier. PP-YOLOE avoids using special operators, such as deformable convolution or matrix nms, to be deployed friendly on various hardware. For more details, please refer to our [report](https://arxiv.org/abs/2203.16250). +PP-YOLOE is an excellent single-stage anchor-free model based on PP-YOLOv2, surpassing a variety of popular YOLO models. PP-YOLOE has a series of models, named s/m/l/x, which are configured through width multiplier and depth multiplier. PP-YOLOE avoids using special operators, such as Deformable Convolution or Matrix NMS, to be deployed friendly on various hardware. For more details, please refer to our [report](https://arxiv.org/abs/2203.16250).
@@ -21,28 +21,31 @@ PP-YOLOE is composed of following methods: - Scalable backbone and neck - [Task Alignment Learning](https://arxiv.org/abs/2108.07755) - Efficient Task-aligned head with [DFL](https://arxiv.org/abs/2006.04388) and [VFL](https://arxiv.org/abs/2008.13367) -- [SiLU activation function](https://arxiv.org/abs/1710.05941) +- [SiLU(Swish) activation function](https://arxiv.org/abs/1710.05941) ## Model Zoo -| Model | GPU number | images/GPU | backbone | input shape | Box APval
0.5:0.95 | Box APtest
0.5:0.95 | Params(M) | FLOPs(G) | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | download | config | -|:------------------------:|:-------:|:----------:|:----------:| :-------:| :------------------: | :-------------------: |:---------:|:--------:| :------------: | :---------------------: | :------: | :------: | -| PP-YOLOE-s | 8 | 32 | cspresnet-s | 640 | 43.0 | 43.2 | 7.93 | 17.36 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](./ppyoloe_crn_s_300e_coco.yml) | -| PP-YOLOE-m | 8 | 28 | cspresnet-m | 640 | 49.0 | 49.1 | 23.43 | 49.91 | 123.4 | 208.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](./ppyoloe_crn_m_300e_coco.yml) | -| PP-YOLOE-l | 8 | 20 | cspresnet-l | 640 | 51.4 | 51.6 | 52.20 | 110.07 | 78.1 | 149.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](./ppyoloe_crn_l_300e_coco.yml) | -| PP-YOLOE-x | 8 | 16 | cspresnet-x | 640 | 52.3 | 52.4 | 98.42 | 206.59 | 45.0 | 95.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](./ppyoloe_crn_x_300e_coco.yml) | +| Model | Epoch | GPU number | images/GPU | backbone | input shape | Box APval
0.5:0.95 | Box APtest
0.5:0.95 | Params(M) | FLOPs(G) | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | download | config | +|:------------------------:|:-------:|:-------:|:--------:|:----------:| :-------:| :------------------: | :-------------------: |:---------:|:--------:|:---------------:| :---------------------: | :------: | :------: | +| PP-YOLOE-s | 400 | 8 | 32 | cspresnet-s | 640 | 43.4 | 43.6 | 7.93 | 17.36 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_400e_coco.pdparams) | [config](./ppyoloe_crn_s_400e_coco.yml) | +| PP-YOLOE-s | 300 | 8 | 32 | cspresnet-s | 640 | 43.0 | 43.2 | 7.93 | 17.36 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](./ppyoloe_crn_s_300e_coco.yml) | +| PP-YOLOE-m | 300 | 8 | 28 | cspresnet-m | 640 | 49.0 | 49.1 | 23.43 | 49.91 | 123.4 | 208.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](./ppyoloe_crn_m_300e_coco.yml) | +| PP-YOLOE-l | 300 | 8 | 20 | cspresnet-l | 640 | 51.4 | 51.6 | 52.20 | 110.07 | 78.1 | 149.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](./ppyoloe_crn_l_300e_coco.yml) | +| PP-YOLOE-x | 300 | 8 | 16 | cspresnet-x | 640 | 52.3 | 52.4 | 98.42 | 206.59 | 45.0 | 95.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](./ppyoloe_crn_x_300e_coco.yml) | + ### Comprehensive Metrics -| Model | AP0.5:0.95 | AP0.5 | AP0.75 | APsmall | APmedium | APlarge | ARsmall | ARmedium | ARlarge | download | config | -|:----------------------:|:---------------:|:----------:|:-------------:| :------------:| :-----------: | :----------: |:------------:|:-------------:|:------------:| :-----: | :-----: | -| PP-YOLOE-s | 43.0 | 59.6 | 47.2 | 26.0 | 47.4 | 58.7 | 45.1 | 70.6 | 81.4 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](./ppyoloe_crn_s_300e_coco.yml)| -| PP-YOLOE-m | 49.0 | 65.9 | 53.8 | 30.9 | 53.5 | 65.3 | 50.9 | 74.4 | 84.7 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](./ppyoloe_crn_m_300e_coco.yml)| -| PP-YOLOE-l | 51.4 | 68.6 | 56.2 | 34.8 | 56.1 | 68.0 | 53.1 | 76.8 | 85.6 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](./ppyoloe_crn_l_300e_coco.yml)| -| PP-YOLOE-x | 52.3 | 69.5 | 56.8 | 35.1 | 57.0 | 68.6 | 55.5 | 76.9 | 85.7 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](./ppyoloe_crn_x_300e_coco.yml)| +| Model | Epoch | AP0.5:0.95 | AP0.5 | AP0.75 | APsmall | APmedium | APlarge | ARsmall | ARmedium | ARlarge | download | config | +|:----------------------:|:-----:|:---------------:|:----------:|:-------------:| :------------:| :-----------: | :----------: |:------------:|:-------------:|:------------:| :-----: | :-----: | +| PP-YOLOE-s | 400 | 43.4 | 60.0 | 47.5 | 25.7 | 47.8 | 59.2 | 43.9 | 70.8 | 81.9 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_400e_coco.pdparams) | [config](./ppyoloe_crn_s_400e_coco.yml)| +| PP-YOLOE-s | 300 | 43.0 | 59.6 | 47.2 | 26.0 | 47.4 | 58.7 | 45.1 | 70.6 | 81.4 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](./ppyoloe_crn_s_300e_coco.yml)| +| PP-YOLOE-m | 300 | 49.0 | 65.9 | 53.8 | 30.9 | 53.5 | 65.3 | 50.9 | 74.4 | 84.7 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](./ppyoloe_crn_m_300e_coco.yml)| +| PP-YOLOE-l | 300 | 51.4 | 68.6 | 56.2 | 34.8 | 56.1 | 68.0 | 53.1 | 76.8 | 85.6 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](./ppyoloe_crn_l_300e_coco.yml)| +| PP-YOLOE-x | 300 | 52.3 | 69.5 | 56.8 | 35.1 | 57.0 | 68.6 | 55.5 | 76.9 | 85.7 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](./ppyoloe_crn_x_300e_coco.yml)| **Notes:** -- PP-YOLOE is trained on COCO train2017 dataset and evaluated on val2017 & test-dev2017 dataset，all the model weights are trained for **300 epoches**. +- PP-YOLOE is trained on COCO train2017 dataset and evaluated on val2017 & test-dev2017 dataset. - The model weights in the table of Comprehensive Metrics are **the same as** that in the original Model Zoo, and evaluated on **val2017**. - PP-YOLOE used 8 GPUs for mixed precision training, if **GPU number** or **mini-batch size** is changed, **learning rate** should be adjusted according to the formula **lrnew = lrdefault * (batch_sizenew * GPU_numbernew) / (batch_sizedefault * GPU_numberdefault)**. - PP-YOLOE inference speed is tesed on single Tesla V100 with batch size as 1, **CUDA 10.2**, **CUDNN 7.6.5**, **TensorRT 6.0.1.8** in TensorRT mode. @@ -221,17 +224,6 @@ Model | AP | AP50 - *Due to the limited time, more verification results will be supplemented in the future. You are also welcome to contribute to PP-YOLOE* -### Feature Models - -The PaddleDetection team provides configs and weights of various feature detection models based on PP-YOLOE, which users can download for use: - -|Scenarios | Related Datasets | Links| -| :--------: | :---------: | :------: | -|Pedestrian Detection | CrowdHuman | [pphuman](../pphuman) | -|Vehicle Detection | BDD100K，UA-DETRAC | [ppvehicle](../ppvehicle) | -|Small Object Detection | VisDrone | [visdrone](../visdrone) | - - ## Appendix Ablation experiments of PP-YOLOE. diff --git a/configs/ppyoloe/README_cn.md b/configs/ppyoloe/README_cn.md index a743a364e5e33eaafcc5f7da77c26ded4bafe6ce..598fc8a2e91b6c3f81d2943a29d5e31e06f0cd89 100644 --- a/configs/ppyoloe/README_cn.md +++ b/configs/ppyoloe/README_cn.md @@ -9,7 +9,7 @@ - [附录](#附录) ## 简介 -PP-YOLOE是基于PP-YOLOv2的卓越的单阶段Anchor-free模型，超越了多种流行的yolo模型。PP-YOLOE有一系列的模型，即s/m/l/x，可以通过width multiplier和depth multiplier配置。PP-YOLOE避免使用诸如deformable convolution或者matrix nms之类的特殊算子，以使其能轻松地部署在多种多样的硬件上。更多细节可以参考我们的[report](https://arxiv.org/abs/2203.16250)。 +PP-YOLOE是基于PP-YOLOv2的卓越的单阶段Anchor-free模型，超越了多种流行的YOLO模型。PP-YOLOE有一系列的模型，即s/m/l/x，可以通过width multiplier和depth multiplier配置。PP-YOLOE避免了使用诸如Deformable Convolution或者Matrix NMS之类的特殊算子，以使其能轻松地部署在多种多样的硬件上。更多细节可以参考我们的[report](https://arxiv.org/abs/2203.16250)。
@@ -21,29 +21,31 @@ PP-YOLOE由以下方法组成 - 可扩展的backbone和neck - [Task Alignment Learning](https://arxiv.org/abs/2108.07755) - Efficient Task-aligned head with [DFL](https://arxiv.org/abs/2006.04388)和[VFL](https://arxiv.org/abs/2008.13367) -- [SiLU激活函数](https://arxiv.org/abs/1710.05941) +- [SiLU(Swish)激活函数](https://arxiv.org/abs/1710.05941) ## 模型库 -| 模型 | GPU个数 | 每GPU图片个数 | 骨干网络 | 输入尺寸 | Box APval
0.5:0.95 | Box APtest
0.5:0.95 | Params(M) | FLOPs(G) | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | 模型下载 | 配置文件 | -|:------------------------:|:-------:|:--------:|:----------:| :-------:| :------------------: | :-------------------: |:---------:|:--------:|:---------------:| :---------------------: | :------: | :------: | -| PP-YOLOE-s | 8 | 32 | cspresnet-s | 640 | 43.0 | 43.2 | 7.93 | 17.36 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](./ppyoloe_crn_s_300e_coco.yml) | -| PP-YOLOE-m | 8 | 28 | cspresnet-m | 640 | 49.0 | 49.1 | 23.43 | 49.91 | 123.4 | 208.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](./ppyoloe_crn_m_300e_coco.yml) | -| PP-YOLOE-l | 8 | 20 | cspresnet-l | 640 | 51.4 | 51.6 | 52.20 | 110.07 | 78.1 | 149.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](./ppyoloe_crn_l_300e_coco.yml) | -| PP-YOLOE-x | 8 | 16 | cspresnet-x | 640 | 52.3 | 52.4 | 98.42 | 206.59 | 45.0 | 95.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](./ppyoloe_crn_x_300e_coco.yml) | +| 模型 | Epoch | GPU个数 | 每GPU图片个数 | 骨干网络 | 输入尺寸 | Box APval
0.5:0.95 | Box APtest
0.5:0.95 | Params(M) | FLOPs(G) | V100 FP32(FPS) | V100 TensorRT FP16(FPS) | 模型下载 | 配置文件 | +|:------------------------:|:-------:|:-------:|:--------:|:----------:| :-------:| :------------------: | :-------------------: |:---------:|:--------:|:---------------:| :---------------------: | :------: | :------: | +| PP-YOLOE-s | 400 | 8 | 32 | cspresnet-s | 640 | 43.4 | 43.6 | 7.93 | 17.36 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_400e_coco.pdparams) | [config](./ppyoloe_crn_s_400e_coco.yml) | +| PP-YOLOE-s | 300 | 8 | 32 | cspresnet-s | 640 | 43.0 | 43.2 | 7.93 | 17.36 | 208.3 | 333.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](./ppyoloe_crn_s_300e_coco.yml) | +| PP-YOLOE-m | 300 | 8 | 28 | cspresnet-m | 640 | 49.0 | 49.1 | 23.43 | 49.91 | 123.4 | 208.3 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](./ppyoloe_crn_m_300e_coco.yml) | +| PP-YOLOE-l | 300 | 8 | 20 | cspresnet-l | 640 | 51.4 | 51.6 | 52.20 | 110.07 | 78.1 | 149.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](./ppyoloe_crn_l_300e_coco.yml) | +| PP-YOLOE-x | 300 | 8 | 16 | cspresnet-x | 640 | 52.3 | 52.4 | 98.42 | 206.59 | 45.0 | 95.2 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](./ppyoloe_crn_x_300e_coco.yml) | ### 综合指标 -| 模型 | AP0.5:0.95 | AP0.5 | AP0.75 | APsmall | APmedium | APlarge | ARsmall | ARmedium | ARlarge | 模型下载 | 配置文件 | -|:----------------------:|:---------------:|:----------:|:-------------:| :------------:| :-----------: | :----------: |:------------:|:-------------:|:------------:| :-----: | :-----: | -| PP-YOLOE-s | 43.0 | 59.6 | 47.2 | 26.0 | 47.4 | 58.7 | 45.1 | 70.6 | 81.4 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](./ppyoloe_crn_s_300e_coco.yml)| -| PP-YOLOE-m | 49.0 | 65.9 | 53.8 | 30.9 | 53.5 | 65.3 | 50.9 | 74.4 | 84.7 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](./ppyoloe_crn_m_300e_coco.yml)| -| PP-YOLOE-l | 51.4 | 68.6 | 56.2 | 34.8 | 56.1 | 68.0 | 53.1 | 76.8 | 85.6 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](./ppyoloe_crn_l_300e_coco.yml)| -| PP-YOLOE-x | 52.3 | 69.5 | 56.8 | 35.1 | 57.0 | 68.6 | 55.5 | 76.9 | 85.7 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](./ppyoloe_crn_x_300e_coco.yml)| +| 模型 | Epoch | AP0.5:0.95 | AP0.5 | AP0.75 | APsmall | APmedium | APlarge | ARsmall | ARmedium | ARlarge | 模型下载 | 配置文件 | +|:----------------------:|:-----:|:---------------:|:----------:|:-------------:| :------------:| :-----------: | :----------: |:------------:|:-------------:|:------------:| :-----: | :-----: | +| PP-YOLOE-s | 400 | 43.4 | 60.0 | 47.5 | 25.7 | 47.8 | 59.2 | 43.9 | 70.8 | 81.9 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_400e_coco.pdparams) | [config](./ppyoloe_crn_s_400e_coco.yml)| +| PP-YOLOE-s | 300 | 43.0 | 59.6 | 47.2 | 26.0 | 47.4 | 58.7 | 45.1 | 70.6 | 81.4 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams) | [config](./ppyoloe_crn_s_300e_coco.yml)| +| PP-YOLOE-m | 300 | 49.0 | 65.9 | 53.8 | 30.9 | 53.5 | 65.3 | 50.9 | 74.4 | 84.7 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_m_300e_coco.pdparams) | [config](./ppyoloe_crn_m_300e_coco.yml)| +| PP-YOLOE-l | 300 | 51.4 | 68.6 | 56.2 | 34.8 | 56.1 | 68.0 | 53.1 | 76.8 | 85.6 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_l_300e_coco.pdparams) | [config](./ppyoloe_crn_l_300e_coco.yml)| +| PP-YOLOE-x | 300 | 52.3 | 69.5 | 56.8 | 35.1 | 57.0 | 68.6 | 55.5 | 76.9 | 85.7 | [model](https://paddledet.bj.bcebos.com/models/ppyoloe_crn_x_300e_coco.pdparams) | [config](./ppyoloe_crn_x_300e_coco.yml)| **注意:** -- PP-YOLOE模型使用COCO数据集中train2017作为训练集，使用val2017和test-dev2017作为测试集，模型权重均为训练**300 epoches**得到的。 +- PP-YOLOE模型使用COCO数据集中train2017作为训练集，使用val2017和test-dev2017作为测试集。 - 综合指标的表格与模型库的表格里的模型权重是**同一个权重**，综合指标是使用**val2017**作为验证精度的。 - PP-YOLOE模型训练过程中使用8 GPUs进行混合精度训练，如果**GPU卡数**或者**batch size**发生了改变，你需要按照公式 **lrnew = lrdefault * (batch_sizenew * GPU_numbernew) / (batch_sizedefault * GPU_numberdefault)** 调整学习率。 - PP-YOLOE模型推理速度测试采用单卡V100，batch size=1进行测试，使用**CUDA 10.2**, **CUDNN 7.6.5**，TensorRT推理速度测试使用**TensorRT 6.0.1.8**。 @@ -224,17 +226,6 @@ CUDA_VISIBLE_DEVICES=0 python deploy/python/infer.py --model_dir=output_inferenc - *由于人力/时间有限，后续将会持续补充更多验证结果，也欢迎各位开源用户贡献，共同优化PP-YOLOE* -### 垂类模型应用 - -PaddleDetection团队提供了基于PP-YOLOE的各种垂类检测模型的配置文件和权重，用户可以下载进行使用： - -| 场景 | 相关数据集 | 链接 | -| :--------: | :---------: | :------: | -| 行人检测 | CrowdHuman | [pphuman](../pphuman) | -| 车辆检测 | BDD100K，UA-DETRAC | [ppvehicle](../ppvehicle) | -| 小目标检测 | VisDrone | [visdrone](../visdrone) | - - ## 附录 PP-YOLOE消融实验 diff --git a/configs/ppyoloe/ppyoloe_crn_s_400e_coco.yml b/configs/ppyoloe/ppyoloe_crn_s_400e_coco.yml new file mode 100644 index 0000000000000000000000000000000000000000..dcc1514c2719a0c916ffcf52260e52ccfdd54ea2 --- /dev/null +++ b/configs/ppyoloe/ppyoloe_crn_s_400e_coco.yml @@ -0,0 +1,50 @@ +_BASE_: [ + '../datasets/coco_detection.yml', + '../runtime.yml', + './_base_/optimizer_300e.yml', + './_base_/ppyoloe_crn.yml', + './_base_/ppyoloe_reader.yml', +] + +log_iter: 100 +snapshot_epoch: 10 +weights: output/ppyoloe_crn_s_400e_coco/model_final + +pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/CSPResNetb_s_pretrained.pdparams +depth_mult: 0.33 +width_mult: 0.50 + +TrainReader: + batch_size: 32 + +LearningRate: + base_lr: 0.04 + + +epoch: 400 +LearningRate: + base_lr: 0.025 + schedulers: + - !CosineDecay + max_epochs: 480 + - !LinearWarmup + start_factor: 0. + epochs: 5 + +OptimizerBuilder: + optimizer: + momentum: 0.9 + type: Momentum + regularizer: + factor: 0.0005 + type: L2 + + +PPYOLOEHead: + static_assigner_epoch: 133 + nms: + name: MultiClassNMS + nms_top_k: 10000 + keep_top_k: 300 + score_threshold: 0.01 + nms_threshold: 0.7