From fdfd2c3334aae2d5eca1ca5df7720af3231d38e4 Mon Sep 17 00:00:00 2001 From: tink2123 Date: Sun, 14 Apr 2019 23:58:42 +0800 Subject: [PATCH] polish readme --- PaddleCV/yolov3/README.md | 64 ++++++++++++++++++++++++------------ PaddleCV/yolov3/README_cn.md | 63 +++++++++++++++++++++++------------ 2 files changed, 85 insertions(+), 42 deletions(-) diff --git a/PaddleCV/yolov3/README.md b/PaddleCV/yolov3/README.md index 8b37aded..0464ae33 100644 --- a/PaddleCV/yolov3/README.md +++ b/PaddleCV/yolov3/README.md @@ -9,7 +9,6 @@ - [Training](#training) - [Evaluation](#evaluation) - [Inference and Visualization](#inference-and-visualization) -- [Appendix](#appendix) ## Installation @@ -45,34 +44,35 @@ Train the model on [MS-COCO dataset](http://cocodataset.org/#download), download cd dataset/coco ./download.sh +The data catalog structure is as follows: + +``` + data/coco/ + ├── annotations + │   ├── instances_train2014.json + │   ├── instances_train2017.json + │   ├── instances_val2014.json + │   ├── instances_val2017.json + | ... + ├── train2017 + │   ├── 000000000009.jpg + │   ├── 000000580008.jpg + | ... + ├── val2017 + │   ├── 000000000139.jpg + │   ├── 000000000285.jpg + | ... + +``` ## Training -After data preparation, one can start the training step by: - - python train.py \ - --model_save_dir=output/ \ - --pretrain=${path_to_pretrain_model} - --data_dir=${path_to_data} - -- Set ```export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7``` to specifiy 8 GPU to train. -- For more help on arguments: - - python train.py --help - -**download the pre-trained model:** This sample provides Resnet-50 pre-trained model which is converted from Caffe. The model fuses the parameters in batch normalization layer. One can download pre-trained model as: - - sh ./weights/download.sh - -Set `pretrain` to load pre-trained model. In addition, this parameter is used to load trained model when finetuning as well. -Please make sure that pre-trained model is downloaded and loaded correctly, otherwise, the loss may be NAN during training. - **Install the [cocoapi](https://github.com/cocodataset/cocoapi):** To train the model, [cocoapi](https://github.com/cocodataset/cocoapi) is needed. Install the cocoapi: git clone https://github.com/cocodataset/cocoapi.git - cd PythonAPI + cd cocoapi/PythonAPI # if cython is not installed pip install Cython # Install into global site-packages @@ -81,6 +81,28 @@ To train the model, [cocoapi](https://github.com/cocodataset/cocoapi) is needed. # not to install the COCO API into global site-packages python2 setup.py install --user +**download the pre-trained model:** This sample provides Resnet-50 pre-trained model which is converted from Caffe. The model fuses the parameters in batch normalization layer. One can download pre-trained model as: + + sh ./weights/download.sh + +Set `pretrain` to load pre-trained model. In addition, this parameter is used to load trained model when finetuning as well. +Please make sure that pre-trained model is downloaded and loaded correctly, otherwise, the loss may be NAN during training. + + +**training:** After data preparation, one can start the training step by: + + python train.py \ + --model_save_dir=output/ \ + --pretrain=${path_to_pretrain_model} + --data_dir=${path_to_data} + +- Set ```export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7``` to specifiy 8 GPU to train. +- For more help on arguments: + + python train.py --help + + + **data reader introduction:** * Data reader is defined in `reader.py` . diff --git a/PaddleCV/yolov3/README_cn.md b/PaddleCV/yolov3/README_cn.md index 2c5c7ebb..2141e369 100644 --- a/PaddleCV/yolov3/README_cn.md +++ b/PaddleCV/yolov3/README_cn.md @@ -9,7 +9,6 @@ - [模型训练](#模型训练) - [模型评估](#模型评估) - [模型推断及可视化](#模型推断及可视化) -- [附录](#附录) ## 安装 @@ -47,34 +46,35 @@ YOLOv3 的网络结构由基础特征提取网络、multi-scale特征融合层 cd dataset/coco ./download.sh +数据目录结构如下： + +``` +data/coco/ +├── annotations +│   ├── instances_train2014.json +│   ├── instances_train2017.json +│   ├── instances_val2014.json +│   ├── instances_val2017.json +| ... +├── train2017 +│   ├── 000000000009.jpg +│   ├── 000000580008.jpg +| ... +├── val2017 +│   ├── 000000000139.jpg +│   ├── 000000000285.jpg +| ... + +``` ## 模型训练 -数据准备完毕后，可以通过如下的方式启动训练： - - python train.py \ - --model_save_dir=output/ \ - --pretrain=${path_to_pretrain_model} - --data_dir=${path_to_data} - -- 通过设置export CUDA\_VISIBLE\_DEVICES=0,1,2,3,4,5,6,7指定8卡GPU训练。 -- 可选参数见： - - python train.py --help - -**下载预训练模型：** 本示例提供darknet53预训练模型，该模型转换自作者提供的darknet53在ImageNet上预训练的权重，采用如下命令下载预训练模型： - - sh ./weights/download.sh - -通过初始化`pretrain` 加载预训练模型。同时在参数微调时也采用该设置加载已训练模型。 -请在训练前确认预训练模型下载与加载正确，否则训练过程中损失可能会出现NAN。 - **安装[cocoapi](https://github.com/cocodataset/cocoapi)：** 训练前需要首先下载[cocoapi](https://github.com/cocodataset/cocoapi)： git clone https://github.com/cocodataset/cocoapi.git - cd PythonAPI + cd cocoapi/PythonAPI # if cython is not installed pip install Cython # Install into global site-packages @@ -83,6 +83,27 @@ YOLOv3 的网络结构由基础特征提取网络、multi-scale特征融合层 # not to install the COCO API into global site-packages python2 setup.py install --user +**下载预训练模型：** 本示例提供darknet53预训练模型，该模型转换自作者提供的darknet53在ImageNet上预训练的权重，采用如下命令下载预训练模型： + + sh ./weights/download.sh + +通过初始化`pretrain` 加载预训练模型。同时在参数微调时也采用该设置加载已训练模型。 +请在训练前确认预训练模型下载与加载正确，否则训练过程中损失可能会出现NAN。 + +**开始训练：** 数据准备完毕后，可以通过如下的方式启动训练： + + python train.py \ + --model_save_dir=output/ \ + --pretrain=${path_to_pretrain_model} + --data_dir=${path_to_data} + +- 通过设置export CUDA\_VISIBLE\_DEVICES=0,1,2,3,4,5,6,7指定8卡GPU训练。 +- 可选参数见： + + python train.py --help + + + **数据读取器说明：** * 数据读取器定义在reader.py中。 -- GitLab