https://github.com/devexcale/juuwazakura
https://github.com/devexcale/juuwazakura
Last synced: 8 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/devexcale/juuwazakura
- Owner: devExcale
- Created: 2024-07-11T20:22:00.000Z (almost 2 years ago)
- Default Branch: master
- Last Pushed: 2024-12-08T14:26:07.000Z (over 1 year ago)
- Last Synced: 2025-01-09T08:27:48.990Z (over 1 year ago)
- Language: Python
- Size: 5.73 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# JuuWazaKura
JuuWazaKura (柔技クラ) is an artificial intelligence model whose scope
is detecting and classifying judo techniques in video clips.
## Pipeline
1. Download competition videos from YouTube
2. Extract segments from competition videos given dataset
3. Include segments in dataset
4. Train model
5. Evaluate model
## Environment Variables
| Name | Description | Default |
|---------------------------------|---------------------------------------------------------------------------------|:-----------------|
| `concurrent_fragment_downloads` | How many fragments to download concurrently, a yt-dlp parameter. | `8` |
| `concurrent_clippers` | How many videos to clip concurrently. | `4` |
| `dataset_source` | Path to the folder containing the csv files. | `dataset/` |
| `dataset_clips` | Path to the folder containing the video clips. | `dataset/clips/` |
| `dataset_include` | Comma-separated list of the YouTube videos to process. Empty list includes all. | `[]` |
| `delete_yt` | Whether to delete the original competition videos after clipping. | `True` |
| `log_levelname` | The logging level name. | `INFO` |