Practical overview
When it fits
Use KAPAO when you can run a Python model locally and need pose or keypoint detections from image or video input. The README says its video demo runs fastest at low resolution with few people in frame.
Who it is for
- Computer-vision developers adding multi-person pose or keypoint detection to an image-analysis workflow.
- Researchers or practitioners who want to use the repository’s Python and PyTorch implementation and inference demos.
Documented uses
- Estimate human poses and keypoints from still images.
- Process video and display or save pose-related visualizations.
- Use pose differences between frames for the simple player-tracking approach shown in the squash demo.
Requirements & limits
The public GitHub repository identifies the project as GPL-3.0 licensed.
- The documented environment uses Python 3.6, Conda, pip requirements, and separately downloaded trained models.
- The README provides both CPU and GPU video examples; its reported speeds are tied to specific demo inputs and hardware.
- The repository describes a Gradio web demo on Hugging Face Spaces for CPU inference on short video clips.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up the KAPAO environment and model files
Prepare the documented Python environment and download trained model files before using the demos.
Before you start
- Anaconda or Miniconda
- Python 3.6
Steps
Create and activate a Conda environment named kapao with Python 3.6.
Clone the KAPAO repository.
From the repository directory, install dependencies with pip install -r requirements.txt.
Run python data/scripts/download_models.py to download trained models.
Expected outcome
The repository dependencies are installed and trained model files have been downloaded for the documented inference demos.
Guide limitations
- This guide covers setup only; it does not verify a successful installation or demo run.
Generate pose visualizations from a static image
Run one of the documented image-demo configurations to generate pose-related visual output.
Before you start
- The KAPAO environment and trained model files set up as described in the repository README.
Steps
From the repository, run python demos/image.py --bbox.
Alternatively, run python demos/image.py --pose --face to generate the documented pose and face visualization.
Expected outcome
The image demo generates visual output using one of the documented bounding-box or pose-and-face configurations.
Guide limitations
- The README lists demo commands but does not specify an input-image argument in these examples.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 7 November 2026
- Original website title
- wmcnally/kapao · Official homepage title · retained in its source language