From 841289915b712a92f5f47aa6e89ce25062f0c6aa Mon Sep 17 00:00:00 2001 From: Tau Date: Wed, 12 Apr 2023 18:31:05 +0800 Subject: [PATCH] [Docs] Refine README (#2207) --- README.md | 4 ++-- README_CN.md | 4 ++-- projects/README.md | 6 ++++++ projects/awesome-mmpose/README.md | 35 +++++++++++++++++++++++++++++++ projects/rtmpose/README.md | 28 ++++++++++++++++++++----- projects/rtmpose/README_CN.md | 13 ++++++++++++ 6 files changed, 81 insertions(+), 9 deletions(-) create mode 100644 projects/awesome-mmpose/README.md diff --git a/README.md b/README.md index 823324c3bd..9f685a1c83 100644 --- a/README.md +++ b/README.md @@ -140,13 +140,13 @@ MMPose v1.0.0 is a major update, including many API and config file changes. Cur | DeepPose (CVPR 2014) | done | | RLE (ICCV 2021) | done | | SoftWingloss (TIP 2021) | | -| VideoPose3D (CVPR 2019) | | +| VideoPose3D (CVPR 2019) | in progress | | Hourglass (ECCV 2016) | done | | LiteHRNet (CVPR 2021) | done | | AdaptiveWingloss (ICCV 2019) | done | | SimpleBaseline2D (ECCV 2018) | done | | PoseWarper (NeurIPS 2019) | | -| SimpleBaseline3D (ICCV 2017) | | +| SimpleBaseline3D (ICCV 2017) | in progress | | HMR (CVPR 2018) | | | UDP (CVPR 2020) | done | | VIPNAS (CVPR 2021) | done | diff --git a/README_CN.md b/README_CN.md index ff5f14c50b..09522b20ae 100644 --- a/README_CN.md +++ b/README_CN.md @@ -138,13 +138,13 @@ MMPose v1.0.0 是一个重大更新，包括了大量的 API 和配置文件的 | DeepPose (CVPR 2014) | done | | RLE (ICCV 2021) | done | | SoftWingloss (TIP 2021) | | -| VideoPose3D (CVPR 2019) | | +| VideoPose3D (CVPR 2019) | in progress | | Hourglass (ECCV 2016) | done | | LiteHRNet (CVPR 2021) | done | | AdaptiveWingloss (ICCV 2019) | done | | SimpleBaseline2D (ECCV 2018) | done | | PoseWarper (NeurIPS 2019) | | -| SimpleBaseline3D (ICCV 2017) | | +| SimpleBaseline3D (ICCV 2017) | in progress | | HMR (CVPR 2018) | | | UDP (CVPR 2020) | done | | VIPNAS (CVPR 2021) | done | diff --git a/projects/README.md b/projects/README.md index 3505f96f41..cca7bd947e 100644 --- a/projects/README.md +++ b/projects/README.md @@ -48,4 +48,10 @@ We also provide some documentation listed below to help you get started:
+- **[📖Awesome MMPose](./awesome-mmpose/)**: A list of Tutorials, Papers, Datasets related to MMPose + +
+ +

+ - **What's next? Join the rank of *MMPose contributors* by creating a new project**! diff --git a/projects/awesome-mmpose/README.md b/projects/awesome-mmpose/README.md new file mode 100644 index 0000000000..62bf38c8c8 --- /dev/null +++ b/projects/awesome-mmpose/README.md @@ -0,0 +1,35 @@ +# Awesome MMPose + +A list of resources related to MMPose. Feel free to contribute! + +## Contents + +- [Tutorials](#tutorials) +- [Papers](#papers) +- [Datasets](#datasets) +- [Projects](#projects) + +## Tutorials + +- [MMPose Tutorial (Chinese)](https://github.com/TommyZihao/MMPose_Tutorials) + MMPose 中文 Jupyter 教程，from 同济子豪兄 +- [OpenMMLab Course](https://github.com/open-mmlab/OpenMMLabCourse) + This repository hosts articles, lectures and tutorials on computer vision and OpenMMLab, helping learners to understand algorithms and master our toolboxes in a systematical way. + +## Papers + +- [\[paper\]](https://arxiv.org/abs/2207.10387) [\[code\]](https://github.com/luminxu/Pose-for-Everything) ECCV 2022, Pose for Everything: Towards Category-Agnostic Pose Estimation +- [\[paper\]](https://arxiv.org/abs/2201.04676) [\[code\]](https://github.com/Sense-X/UniFormer) ICLR 2022, UniFormer: Unified Transformer for Efficient Spatiotemporal Representation Learning +- [\[paper\]](https://arxiv.org/abs/2201.07412) [\[code\]](https://github.com/aim-uofa/Poseur) ECCV 2022, Poseur:Direct Human Pose Regression with Transformers +- [\[paper\]](https://arxiv.org/abs/2106.03348) [\[code\]](https://github.com/ViTAE-Transformer/ViTAE-Transformer) NeurIPS 2022, ViTAEv2: Vision Transformer Advanced by Exploring Inductive Bias for Image Recognition and Beyond +- [\[paper\]](https://arxiv.org/abs/2204.10762) [\[code\]](https://github.com/ZiyiZhang27/Dite-HRNet) IJCAI-ECAI 2021, Dite-HRNet:Dynamic Lightweight High-Resolution Network for Human Pose Estimation +- [\[paper\]](https://arxiv.org/abs/2302.08453) [\[code\]](https://github.com/TencentARC/T2I-Adapter) T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models +- [\[paper\]](https://arxiv.org/pdf/2303.11638.pdf) [\[code\]](https://github.com/Gengzigang/PCT) CVPR 2023, Human Pose as Compositional Tokens + +## Datasets + +Waiting for your contribution! + +## Projects + +Waiting for your contribution! diff --git a/projects/rtmpose/README.md b/projects/rtmpose/README.md index 01f454dfdb..528895e3a2 100644 --- a/projects/rtmpose/README.md +++ b/projects/rtmpose/README.md @@ -267,10 +267,12 @@ bash opencv.sh # Compile executable programs bash build.sh -# inference for an image +# Inference for an image +# Please pass the folder of the model, not the model file ./bin/det_pose {det work-dir} {pose work-dir} {your_img.jpg} --device cpu -# inference for a video +# Inference for a video +# Please pass the folder of the model, not the model file ./bin/pose_tracker {det work-dir} {pose work-dir} {your_video.mp4} --device cpu ``` @@ -296,10 +298,12 @@ bash opencv.sh # Compile executable programs bash build.sh -# inference for an image +# Inference for an image +# Please pass the folder of the model, not the model file ./bin/det_pose {det work-dir} {pose work-dir} {your_img.jpg} --device cuda -# inference for a video +# Inference for a video +# Please pass the folder of the model, not the model file ./bin/pose_tracker {det work-dir} {pose work-dir} {your_video.mp4} --device cuda ``` @@ -312,11 +316,21 @@ For details, see [Pipeline Inference](#-step4-pipeline-inference). 1. Download the [pre-compiled SDK](https://github.com/open-mmlab/mmdeploy/releases). 2. Unzip the SDK and go to the `sdk/python` folder. 3. Install `mmdeploy_python` via `.whl` file. + +```shell +pip install {file_name}.whl +``` + 4. Download the [sdk models](https://download.openmmlab.com/mmpose/v1/projects/rtmpose/rtmpose-cpu.zip) and unzip. 5. Inference with `pose_tracker.py`: +> Note: + +- If you meet `ImportError: DLL load failed while importing mmdeploy_python`, please copy `thirdparty/onnxruntime/lib/onnxruntime.dll` to `site-packages/mmdeploy_python/` of your current Python env. + ```shell # go to ./sdk/example/python +# Please pass the folder of the model, not the model file python pose_tracker.py cpu {det work-dir} {pose work-dir} {your_video.mp4} ``` @@ -363,7 +377,11 @@ example\cpp\build\Release ### MMPose demo scripts -MMPose provides demo scripts to conduct [inference with existing models](https://mmpose.readthedocs.io/en/1.x/user_guides/inference.html). +MMPose provides demo scripts to conduct [inference with existing models](https://mmpose.readthedocs.io/en/latest/user_guides/inference.html). + +**Note:** + +- Inferencing with Pytorch can not reach the maximum speed of RTMPose, just for verification. ```shell # go to the mmpose folder diff --git a/projects/rtmpose/README_CN.md b/projects/rtmpose/README_CN.md index dfad499f18..041c05e2b0 100644 --- a/projects/rtmpose/README_CN.md +++ b/projects/rtmpose/README_CN.md @@ -262,9 +262,11 @@ bash opencv.sh bash build.sh # 图片推理 +# 请传入模型目录，而不是模型文件 ./bin/det_pose {det work-dir} {pose work-dir} {your_img.jpg} --device cpu # 视频推理 +# 请传入模型目录，而不是模型文件 ./bin/pose_tracker {det work-dir} {pose work-dir} {your_video.mp4} --device cpu ``` @@ -290,9 +292,11 @@ bash opencv.sh bash build.sh # 图片推理 +# 请传入模型目录，而不是模型文件 ./bin/det_pose {det work-dir} {pose work-dir} {your_img.jpg} --device cuda # 视频推理 +# 请传入模型目录，而不是模型文件 ./bin/pose_tracker {det work-dir} {pose work-dir} {your_video.mp4} --device cuda ``` @@ -313,8 +317,13 @@ pip install {file_name}.whl 4. 下载 [sdk 模型](https://download.openmmlab.com/mmpose/v1/projects/rtmpose/rtmpose-cpu.zip)并解压。 5. 使用 `pose_tracker.py` 进行推理： +**提示：** + +- 如果遇到 `ImportError: DLL load failed while importing mmdeploy_python`，请复制 `thirdparty/onnxruntime/lib/onnxruntime.dll` 到当前环境中 python 安装目录的 `site-packages/mmdeploy_python/`。 + ```shell # 进入 ./sdk/example/python +# 请传入模型目录，而不是模型文件 python pose_tracker.py cpu {det work-dir} {pose work-dir} {your_video.mp4} ``` @@ -363,6 +372,10 @@ example\cpp\build\Release 通过 MMPose 提供的 demo 脚本可以基于 Pytorch 快速进行[模型推理](https://mmpose.readthedocs.io/en/latest/user_guides/inference.html)和效果验证。 +**提示：** + +- 基于 Pytorch 推理并不能达到 RTMPose 模型的真实推理速度，只用于模型效果验证。 + ```shell # 前往 mmpose 目录 cd ${PATH_TO_MMPOSE}