| Build Type | Linux |
MacOS |
Windows |
|---|---|---|---|
| Build Status |  |
|
OpenPose has represented the first real-time multi-person system to jointly detect human body, hand, facial, and foot keypoints (in total 135 keypoints) on single images.
It is authored by Ginés Hidalgo, Zhe Cao, Tomas Simon, Shih-En Wei, Yaadhav Raaj, Hanbyul Joo, and Yaser Sheikh. It is maintained by Ginés Hidalgo and Yaadhav Raaj. OpenPose would not be possible without the CMU Panoptic Studio dataset. We would also like to thank all the people who has helped OpenPose in any way.
Authors Ginés Hidalgo (left) and Hanbyul Joo (right) in front of the CMU Panoptic Studio
Contents
Results
Whole-body (Body, Foot, Face, and Hands) 2D Pose Estimation
Testing OpenPose: (Left) Crazy Uptown Funk flashmob in Sydney video sequence. (Center and right) Authors Ginés Hidalgo and Tomas Simon testing face and hands
Whole-body 3D Pose Reconstruction and Estimation
Tianyi Zhao testing the OpenPose 3D Module
Unity Plugin
Tianyi Zhao and Ginés Hidalgo testing the OpenPose Unity Plugin
Runtime Analysis
We show an inference time comparison between the 3 available pose estimation libraries (same hardware and conditions): OpenPose, Alpha-Pose (fast Pytorch version), and Mask R-CNN. The OpenPose runtime is constant, while the runtime of Alpha-Pose and Mask R-CNN grow linearly with the number of people. More details here.
Features
- Main Functionality:
- 2D real-time multi-person keypoint detection:
- 15, 18 or 25-keypoint body/foot keypoint estimation, including 6 foot keypoints. Runtime invariant to number of detected people.
- 2x21-keypoint hand keypoint estimation. Runtime depends on number of detected people. See OpenPose Training for a runtime invariant alternative.
- 70-keypoint face keypoint estimation. Runtime depends on number of detected people. See OpenPose Training for a runtime invariant alternative.
- 3D real-time single-person keypoint detection:
- 3D triangulation from multiple single views.
- Synchronization of Flir cameras handled.
- Compatible with Flir/Point Grey cameras.
- Calibration toolbox: Estimation of distortion, intrinsic, and extrinsic camera parameters.
- Single-person tracking for further speedup or visual smoothing.
- 2D real-time multi-person keypoint detection:
- Input: Image, video, webcam, Flir/Point Grey, IP camera, and support to add your own custom input source (e.g., depth camera).
- Output: Basic image + keypoint display/saving (PNG, JPG, AVI, ...), keypoint saving (JSON, XML, YML, ...), keypoints as array class, and support to add your own custom output code (e.g., some fancy UI).
- OS: Ubuntu (20, 18, 16, 14), Windows (10, 8), Mac OSX, Nvidia TX2.
- Hardware compatibility: CUDA (Nvidia GPU), OpenCL (AMD GPU), and non-GPU (CPU-only) versions.
- Usage Alternatives:
- Command-line demo for built-in functionality.
- C++ API and Python API for custom functionality. E.g., adding your custom inputs, pre-processing, post-posprocessing, and output steps.
For further details, check the major released features and release notes docs.
Related Work
- OpenPose training code
- OpenPose foot dataset
- OpenPose Unity Plugin
- OpenPose papers published in IEEE TPAMI and CVPR. Cite them in your publications if OpenPose helps your research! (Links and more details in the Citation section below).
Installation
If you want to use OpenPose without installing or writing any code, simply download and use the latest Windows portable version of OpenPose!
Otherwise, you could build OpenPose from source. See the installation doc for all the alternatives.
Quick Start Overview
Simply use the OpenPose Demo from your favorite command-line tool (e.g., Windows PowerShell or Ubuntu Terminal). E.g., this example runs OpenPose on your webcam and displays the body keypoints:
# Ubuntu
./build/examples/openpose/openpose.bin
:: Windows - Portable Demo
bin\OpenPoseDemo.exe --video examples\media\video.avi
You can also add any of the available flags in any order. E.g., the following example runs on a video (--video {PATH}), enables face (--face) and hands (--hand), and saves the output keypoints on JSON files on disk (--write_json {PATH}).
# Ubuntu
./build/examples/openpose/openpose.bin --video examples/media/video.avi --face --hand --write_json output_json_folder/
:: Windows - Portable Demo
bin\OpenPoseDemo.exe --video examples\media\video.avi --face --hand --write_json output_json_folder/
Optionally, you can also extend OpenPose's functionality from its Python and C++ APIs. After installing OpenPose, check its official doc for a quick overview of all the alternatives and tutorials.
Send Us Feedback!
Our library is open source for research purposes, and we want to improve it! So let us know (create a new GitHub issue or pull request, email us, etc.) if you...
- Find/fix any bug (in functionality or speed) or know how to speed up or improve any part of OpenPose.
- Want to add/show some cool functionality/demo/project made on top of OpenPose. We can add your project link to our Community-based Projects section or even integrate it with OpenPose!
Citation
Please cite these papers in your publications if OpenPose helps your research. All of OpenPose is based on OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields, while the hand and face detectors also use Hand Keypoint Detection in Single Images using Multiview Bootstrapping (the face detector was trained using the same procedure than the hand detector).
@article{8765346,
author = {Z. {Cao} and G. {Hidalgo Martinez} and T. {Simon} and S. {Wei} and Y. A. {Sheikh}},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
title = {OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields},
year = {2019}
}
@inproceedings{simon2017hand,
author = {Tomas Simon and Hanbyul Joo and Iain Matthews and Yaser Sheikh},
booktitle = {CVPR},
title = {Hand Keypoint Detection in Single Images using Multiview Bootstrapping},
year = {2017}
}
@inproceedings{cao2017realtime,
author = {Zhe Cao and Tomas Simon and Shih-En Wei and Yaser Sheikh},
booktitle = {CVPR},
title = {Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields},
year = {2017}
}
@inproceedings{wei2016cpm,
author = {Shih-En Wei and Varun Ramakrishna and Takeo Kanade and Yaser Sheikh},
booktitle = {CVPR},
title = {Convolutional pose machines},
year = {2016}
}
Paper links:
- OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields:
- Hand Keypoint Detection in Single Images using Multiview Bootstrapping
- Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- Convolutional Pose Machines
License
OpenPose is freely available for free non-commercial use, and may be redistributed under these conditions. Please, see the license for further details. Interested in a commercial license? Check this FlintBox link. For commercial queries, use the Contact section from the FlintBox link and also send a copy of that message to Yaser Sheikh.
![[libprotobuf ERROR google/protobuf/message_lite.cc:123] Can't parse message of type](https://avatars.githubusercontent.com/u/26670030?v=4)
59 Dec 06, 2022
27 Oct 28, 2022
2.4k Jan 06, 2023
75 Dec 27, 2022
81 Oct 06, 2022
27 Nov 17, 2022
44 Dec 20, 2022
170 Dec 15, 2022
132 Nov 21, 2022
72 Dec 07, 2022
50 Sep 20, 2022
1 Jan 26, 2022
1 Jan 09, 2022
3 Dec 13, 2021
1 Nov 17, 2021
103 Jan 06, 2023
65 Dec 22, 2022
624 Jan 03, 2023
33 Nov 30, 2022
53 Dec 09, 2022