[doc] Update changelog and doc (#5816)

* update doc, test=document_fix * update doc, test=document_fix * update changelog, test=document_fix * update readme model, test=document_fix

[doc] Update changelog and doc (#5816)
* update doc, test=document_fix * update doc, test=document_fix * update changelog, test=document_fix * update readme model, test=document_fix
c1fe1dd2 · Feng Ni · GitHub · 97e9e75b · c1fe1dd2 · c1fe1dd2
8 changed file
--- a/README_cn.md
+++ b/README_cn.md
@@ -31,6 +31,8 @@
        - 应用落地难点剖析与解决方案
        - 行人分析实战与Docker云上训练部署

+  🔥 **[课程回放链接](https://aistudio.baidu.com/aistudio/education/group/info/23670)**🔥
+
    赶紧扫码报名上车吧！!

    <div align="left">
@@ -42,9 +44,10 @@

 - 🔥 **2022.3.24：PaddleDetection发布[release/2.4版本](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.4)**

-  - 发布高精度云边一体SOTA目标检测模型[PP-YOLOE](configs/ppyoloe)，COCO数据集精度51.4%，V100预测速度78.1 FPS，支持混合精度训练，训练较PP-YOLOv2加速33%，全系列多尺度模型，满足不同硬件算力需求，可适配服务器、边缘端GPU及其他服务器端AI加速卡。
+  - 发布高精度云边一体SOTA目标检测模型[PP-YOLOE](configs/ppyoloe)，发布s/m/l/x版本，l版本COCO test2017数据集精度51.4%，V100预测速度78.1 FPS，支持混合精度训练，训练较PP-YOLOv2加速33%，全系列多尺度模型，满足不同硬件算力需求，可适配服务器、边缘端GPU及其他服务器端AI加速卡。
  - 发布边缘端和CPU端超轻量SOTA目标检测模型[PP-PicoDet增强版](configs/picodet)，精度提升2%左右，CPU预测速度提升63%，新增参数量0.7M的PicoDet-XS模型，提供模型稀疏化和量化功能，便于模型加速，各类硬件无需单独开发后处理模块，降低部署门槛。
  - 发布实时行人分析工具[PP-Human](deploy/pphuman)，支持行人跟踪、人流量统计、人体属性识别与摔倒检测四大能力，基于真实场景数据特殊优化，精准识别各类摔倒姿势，适应不同环境背景、光线及摄像角度。
+  - 新增[YOLOX](configs/yolox)目标检测模型，支持nano/tiny/s/m/l/x版本，x版本COCO val2017数据集精度51.8%。

 - 2021.11.03: PaddleDetection发布[release/2.3版本](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.3)

@@ -135,6 +138,8 @@
            <li>YOLOv4</li>  
            <li>PP-YOLOv1/v2</li>
            <li>PP-YOLO-Tiny</li>
+            <li>PP-YOLOE</li>
+            <li>YOLOX</li>
            <li>SSD</li>
            <li>CornerNet-Squeeze</li>
            <li>FCOS</li>  
@@ -160,7 +165,7 @@
        <ul>
            <li>JDE</li>
            <li>FairMOT</li>
-            <li>DeepSort</li>
+            <li>DeepSORT</li>
        </ul>
        <li><b>KeyPoint-Detection</b></li>
        <ul>
@@ -240,6 +245,7 @@
          <li>Color Distort</li>  
          <li>Random Erasing</li>  
          <li>Mixup </li>
+          <li>AugmentHSV</li>
          <li>Mosaic</li>
          <li>Cutmix </li>
          <li>Grid Mask</li>
@@ -344,6 +350,7 @@
  - [DeepSORT](configs/mot/deepsort/README_cn.md)
  - [JDE](configs/mot/jde/README_cn.md)
  - [FairMOT](configs/mot/fairmot/README_cn.md)
+  - [ByteTrack](configs/mot/bytetrack/README.md)
 - 垂类领域
  - [行人检测](configs/pedestrian/README.md)
  - [车辆检测](configs/vehicle/README.md)

--- a/README_en.md
+++ b/README_en.md
@@ -18,9 +18,10 @@ English | [简体中文](README_cn.md)

 - 🔥 **2022.3.24：PaddleDetection [release 2.4 version](https://github.com/PaddlePaddle/PaddleDetection/tree/release/2.4)**

-  - Release GPU SOTA object detection series models (s/m/l/x) [PP-YOLOE](configs/ppyoloe), achieving mAP as 51.4% on COCO test dataset and 78.1 FPS on Nvidia V100, supporting AMP training and its training speed is 33% faster than PP-YOLOv2.
+  - Release GPU SOTA object detection series models (s/m/l/x) [PP-YOLOE](configs/ppyoloe), supporting s/m/l/x version, achieving mAP as 51.4% on COCO test dataset and 78.1 FPS on Nvidia V100 by PP-YOLOE-l, supporting AMP training and its training speed is 33% faster than PP-YOLOv2.
  - Release enhanced models of [PP-PicoDet](configs/picodet), including PP-PicoDet-XS model with 0.7M parameters, its mAP promoted ~2% on COCO, inference speed accelerated 63% on CPU, and post-processing integrated into the network to optimize deployment pipeline.
  - Release real-time human analysis tool [PP-Human](deploy/pphuman), which is based on data from real-life situations, supporting pedestrian detection, attribute recognition, human tracking, multi-camera tracking, human statistics and action recognition.
+  - Release [YOLOX](configs/yolox), supporting nano/tiny/s/m/l/x version, achieving mAP as 51.8% on COCO val dataset by YOLOX-x.

 - 2021.11.03: Release [release/2.3](https://github.com/PaddlePaddle/Paddleetection/tree/release/2.3) version. Release mobile object detection model ⚡[PP-PicoDet](configs/picodet), mobile keypoint detection model ⚡[PP-TinyPose](configs/keypoint/tiny_pose)，Real-time tracking system [PP-Tracking](deploy/pptracking). Release object detection models, including [Swin-Transformer](configs/faster_rcnn), [TOOD](configs/tood), [GFL](configs/gfl), release [Sniper](configs/sniper) tiny object detection models and optimized [PP-YOLO-EB](configs/ppyolo) model for EdgeBoard. Release mobile keypoint detection model [Lite HRNet](configs/keypoint).

@@ -107,6 +108,8 @@ PaddleDetection is an end-to-end object detection development kit based on Paddl
            <li>YOLOv4</li>  
            <li>PP-YOLOv1/v2</li>
            <li>PP-YOLO-Tiny</li>
+            <li>PP-YOLOE</li>
+            <li>YOLOX</li>
            <li>SSD</li>
            <li>CornerNet-Squeeze</li>
            <li>FCOS</li>  
@@ -132,7 +135,7 @@ PaddleDetection is an end-to-end object detection development kit based on Paddl
        <ul>
            <li>JDE</li>
            <li>FairMOT</li>
-            <li>DeepSort</li>
+            <li>DeepSORT</li>
        </ul>
        <li><b>KeyPoint-Detection</b></li>
        <ul>
@@ -213,6 +216,7 @@ PaddleDetection is an end-to-end object detection development kit based on Paddl
          <li>Random Erasing</li>  
          <li>Mixup </li>
          <li>Mosaic</li>
+          <li>AugmentHSV</li>
          <li>Cutmix </li>
          <li>Grid Mask</li>
          <li>Auto Augment</li>  
@@ -319,6 +323,7 @@ The relationship between COCO mAP and FPS on Qualcomm Snapdragon 865 of represen
  - [DeepSORT](configs/mot/deepsort/README.md)
  - [JDE](configs/mot/jde/README.md)
  - [FairMOT](configs/mot/fairmot/README.md)
+  - [ByteTrack](configs/mot/bytetrack/README.md)
 - Vertical field
  - [Face detection](configs/face_detection/README_en.md)
  - [Pedestrian detection](configs/pedestrian/README.md)

--- a/configs/ppyoloe/README_cn.md
+++ b/configs/ppyoloe/README_cn.md
@@ -37,7 +37,7 @@ PP-YOLOE由以下方法组成
 - 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 **lr<sub>new</sub> = lr<sub>default</sub> * (batch_size<sub>new</sub> * GPU_number<sub>new</sub>) / (batch_size<sub>default</sub> * GPU_number<sub>default</sub>)** 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**。
 - 参考[速度测试](#速度测试)以复现PP-YOLOE推理速度测试结果。
- 如果你设置了`--run_benchnark=True`, 你首先需要安装以下依赖`pip install pynvml psutil GPUtil`。
+- 如果你设置了`--run_benchmark=True`, 你首先需要安装以下依赖`pip install pynvml psutil GPUtil`。

 ## 使用教程


--- a/configs/yolox/README.md
+++ b/configs/yolox/README.md
@@ -10,10 +10,11 @@
 | YOLOX-s        |  640     |    8      |   300e    |     ----    |  40.4  | [下载链接](https://paddledet.bj.bcebos.com/models/yolox_s_300e_coco.pdparams) | [配置文件](./yolox_s_300e_coco.yml) |
 | YOLOX-m        |  640     |    8      |   300e    |     ----    |  46.9  | [下载链接](https://paddledet.bj.bcebos.com/models/yolox_m_300e_coco.pdparams) | [配置文件](./yolox_m_300e_coco.yml) |
 | YOLOX-l        |  640     |    8      |   300e    |     ----    |  50.1  | [下载链接](https://paddledet.bj.bcebos.com/models/yolox_l_300e_coco.pdparams) | [配置文件](./yolox_l_300e_coco.yml) |
-| YOLOX-x        |  640     |    8      |   300e    |     ----    |  51.4  | [下载链接](https://paddledet.bj.bcebos.com/models/yolox_x_300e_coco.pdparams) | [配置文件](./yolox_x_300e_coco.yml) |
+| YOLOX-x        |  640     |    8      |   300e    |     ----    |  51.8  | [下载链接](https://paddledet.bj.bcebos.com/models/yolox_x_300e_coco.pdparams) | [配置文件](./yolox_x_300e_coco.yml) |

 **注意:**
-  - 以上模型默认采用8 GPUs训练，总batch_size为64，均训练300 epochs；
+  - YOLOX模型训练使用COCO train2017作为训练集，Box AP为在COCO val2017上的`mAP(IoU=0.5:0.95)`结果；
+  - YOLOX模型训练过程中默认使用8 GPUs进行混合精度训练，默认单卡batch_size为8，如果**GPU卡数**或者**batch size**发生了改变，你需要按照公式 **lr<sub>new</sub> = lr<sub>default</sub> * (batch_size<sub>new</sub> * GPU_number<sub>new</sub>) / (batch_size<sub>default</sub> * GPU_number<sub>default</sub>)** 调整学习率；
  - 为保持高mAP的同时提高推理速度，可以将[yolox_cspdarknet.yml](_base_/yolox_cspdarknet.yml)中的`nms_top_k`修改为`1000`，将`keep_top_k`修改为`100`，mAP会下降约0.1~0.2%；
  - 为快速的demo演示效果，可以将[yolox_cspdarknet.yml](_base_/yolox_cspdarknet.yml)中的`score_threshold`修改为`0.25`，将`nms_threshold`修改为`0.45`，但mAP会下降较多；

@@ -45,15 +46,19 @@ CUDA_VISIBLE_DEVICES=0 python tools/infer.py -c configs/yolox/yolox_s_300e_coco.
 ```

 ### 4. 部署
+
+#### 4.1. 导出模型
 YOLOX在GPU上推理部署或benchmark测速等需要通过`tools/export_model.py`导出模型。
 运行以下的命令进行导出：
 ```bash
 python tools/export_model.py -c configs/yolox/yolox_s_300e_coco.yml -o weights=https://paddledet.bj.bcebos.com/models/yolox_s_300e_coco.pdparams
 ```
-`deploy/python/infer.py`使用上述导出后的Paddle Inference模型用于推理和benchnark测速.
+
+#### 4.2. Python部署
+`deploy/python/infer.py`使用上述导出后的Paddle Inference模型用于推理和benchnark测速，如果设置了`--run_benchmark=True`, 首先需要安装以下依赖`pip install pynvml psutil GPUtil`。

 ```bash
-# 推理单张图片
+# Python部署推理单张图片
 python deploy/python/infer.py --model_dir=output_inference/yolox_s_300e_coco --image_file=demo/000000014439_640x640.jpg --device=gpu

 # 推理文件夹下的所有图片
@@ -69,6 +74,12 @@ python deploy/python/infer.py --model_dir=output_inference/yolox_s_300e_coco --i
 python deploy/python/infer.py --model_dir=output_inference/yolox_s_300e_coco --image_file=demo/000000014439_640x640.jpg --device=gpu --run_benchmark=True --trt_max_shape=640 --trt_min_shape=640 --trt_opt_shape=640 --run_mode=trt_fp16
 ```

+#### 4.2. C++部署
+`deploy/cpp/build/main`使用上述导出后的Paddle Inference模型用于C++推理部署, 首先按照[docs](../../deploy/cpp/docs)编译安装环境。
+```bash
+# C++部署推理单张图片
+./deploy/cpp/build/main --model_dir=output_inference/yolox_s_300e_coco/ --image_file=demo/000000014439_640x640.jpg --run_mode=paddle --device=GPU --threshold=0.5 --output_dir=cpp_infer_output/yolox_s_300e_coco
+```

 ## Citations
 ```

--- a/docs/CHANGELOG.md
+++ b/docs/CHANGELOG.md
@@ -7,7 +7,7 @@
 ### 2.4(03.24/2022)

 - PP-YOLOE：
-  - 发布PP-YOLOE特色模型，COCO数据集精度51.4%，V100预测速度78.1 FPS，精度速度服务器端SOTA
+  - 发布PP-YOLOE特色模型，l版本COCO test2017数据集精度51.4%，V100预测速度78.1 FPS，精度速度服务器端SOTA
  - 发布s/m/l/x系列模型，打通TensorRT、ONNX部署能力
  - 支持混合精度训练，训练较PP-YOLOv2加速33%

@@ -22,6 +22,9 @@
  - ReID支持Centroid模型
  - 动作识别支持ST-GCN摔倒检测

+- 模型丰富度:
+  - 发布YOLOX，支持nano/tiny/s/m/l/x版本，x版本COCO val2017数据集精度51.8%
+
 - 框架功能优化：
  - EMA训练速度优化20%，优化EMA训练模型保存方式
  - 支持infer预测结果保存为COCO格式

--- a/docs/CHANGELOG_en.md
+++ b/docs/CHANGELOG_en.md
@@ -7,14 +7,14 @@ English | [简体中文](./CHANGELOG.md)
 ### 2.4(03.24/2022)

 - PP-YOLOE：
-  - Release PP-YOLOE object detection models, achieve mAP as 51.4% on COCO test dataset and 78.1 FPS on Nvidia V100, reach SOTA performance for object detection on GPU``
+  - Release PP-YOLOE object detection models, achieve mAP as 51.4% on COCO test dataset and 78.1 FPS on Nvidia V100 by PP-YOLOE-l, reach SOTA performance for object detection on GPU``
  - Release series models: s/m/l/x, and support deployment base on TensorRT & ONNX
  - Spport AMP training and training speed is 33% faster than PP-YOLOv2

 - PP-PicoDet:
  - Release enhanced models of PP-PicoDet, mAP promoted ~2% on COCO and inference speed accelerated 63% on CPU
  - Release PP-PicoDet-XS model with 0.7M parameters
-  - Post-processing integrated into the network to optimize deployment pipeline 
+  - Post-processing integrated into the network to optimize deployment pipeline

 - PP-Human：
  - Release PP-Human human analysis pipeline，including pedestrian detection, attribute recognition, human tracking, multi-camera tracking, human statistics, action recognition. Supporting deployment with TensorRT
@@ -22,6 +22,9 @@ English | [简体中文](./CHANGELOG.md)
  - Release Centroid model for ReID
  - Release ST-GCN model for falldown action recognition

+- Model richness:
+  - Publish YOLOX object detection model, release series models: nano/tiny/s/m/l/x, and YOLOX-x achieves mAP as 51.8% on COCO val2017 dataset
+
 - Function Optimize：
  - Optimize 20% training speed when training with EMA, improve saving method of EMA weights
  - Support saving inference results in COCO format

--- a/docs/MODEL_ZOO_cn.md
+++ b/docs/MODEL_ZOO_cn.md
@@ -87,6 +87,14 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型

 请参考[PicoDet](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/picodet)

+### PP-YOLOE
+
+请参考[PP-YOLOE](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe)
+
+### YOLOX
+
+请参考[YOLOX](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/yolox)
+

 ## 旋转框检测

@@ -94,6 +102,7 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型

 请参考[S2ANet](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/dota/)

+
 ## 关键点检测

 ### PP-TinyPose
@@ -108,16 +117,21 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型

 请参考[HigherHRNet](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/keypoint/higherhrnet)

+
 ## 多目标跟踪

-### DeepSort
+### DeepSORT

-请参考[DeepSort](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/deepsort)
+请参考[DeepSORT](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/deepsort)

 ### JDE

 请参考[JDE](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/jde)

-### fairmot 
+### FairMOT
+
+请参考[FairMOT](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/fairmot)
+
+### ByteTrack

-请参考[FairMot](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/fairmot)
+请参考[ByteTrack](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/bytetrack)
--- a/docs/MODEL_ZOO_en.md
+++ b/docs/MODEL_ZOO_en.md
@@ -86,6 +86,14 @@ Please refer to[GFL](https://github.com/PaddlePaddle/PaddleDetection/tree/develo

 Please refer to[PicoDet](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/picodet)

+### PP-YOLOE
+
+Please refer to[PP-YOLOE](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe)
+
+### YOLOX
+
+Please refer to[YOLOX](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/yolox)
+

 ## Rotating frame detection

@@ -93,6 +101,7 @@ Please refer to[PicoDet](https://github.com/PaddlePaddle/PaddleDetection/tree/de

 Please refer to[S2ANet](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/dota/)

+
 ## KeyPoint Detection

 ### PP-TinyPose
@@ -107,16 +116,21 @@ Please refer to [HRNet](https://github.com/PaddlePaddle/PaddleDetection/tree/dev

 Please refer to [HigherHRNet](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/keypoint/higherhrnet)

+
 ## Multi-Object Tracking

-### DeepSort
+### DeepSORT

-Please refer to [DeepSort](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/deepsort)
+Please refer to [DeepSORT](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/deepsort)

 ### JDE

 Please refer to [JDE](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/jde)

-### fairmot 
+### FairMOT
+
+Please refer to [FairMOT](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/fairmot)
+
+### ByteTrack

-Please refer to [FairMot](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/fairmot)
\ No newline at end of file
+Please refer to [ByteTrack](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/mot/bytetrack)