Your go-to tool for extracting Estimated Pose Keypoints from Videos in-the-wild!

Source

It's a command-line tool which you can use to run the project. The intention behind creating this project is to enable Developers, Researchers, and Animators to extract Key Point information from raw in-the-wild videos. Further, the dataset extracted can be used for future applications.

As of now we are just utilizing 1 model for pose-estimation model.

• [https://docs.ultralytics.com/tasks/pose/] YoloV8-Pose

The tool can be used for the following use cases:

• Getting information about movements from detected persons
• Creation of datasets for Human Motion Synthesis/Pose Estimation 3D
• Animators requiring Pose Reference videos
• Any other use case you can suggest

The tool usage is simple. Just trigger the following commands from the command line tool.

# Install the virtual environment dependency
pip install virtualenv

# Replace "myenv" with the name you want to give to your virtual environment
virtualenv myenv

# Activate the environment
source myenv/bin/activate

#Install the necessary dependencies
pip install -r requirements.txt

python localrunner.py --YT_LIST_FILE yt_list.csv --UPLOAD_FOLDER local_uploads

python localrunner.py --YT_LIST_FILE yt_list.csv --UPLOAD_FOLDER local_uploads --is_keypoint_file

python localrunner.py --YT_LIST_FILE yt_list.csv --UPLOAD_FOLDER local_uploads --is_keypoint_file --is_keypoint_video

python localrunner.py --YT_LIST_FILE yt_list.csv --UPLOAD_FOLDER local_uploads --is_keypoint_file --is_keypoint_video --keypoint_bones

As of now, the tool is still under development phase and has the following limitations:

• No documentation for usage
• No support for diverse models
• No pre-processing models/algorithms (Like background removal, upscaling/downscaling, etc...)
• No support for multi-processing (Still under development as of yet)
• Limited to only 2D point estimation (Current model provides a Visible output for z-axis)

For future developments, we are still working on bringing up the limitations mentioned above to make the tool diverse.

Feel free to raise PR if you can improve the project further. We do require help for the following:

• Support for Multi-Processing
• Model Zoo
• Code Refactoring
• Unit Tests/Integration Tests
• Web-UI for interactive sessions