OSL-ActionSpotting is a plug-and-play library that unifies action spotting algorithms.
- A technical report of this library will be provided soon.
- Support SoTA TAD methods with modular design. We decompose the TAD pipeline into different components, and implement them in a modular way. This design makes it easy to implement new methods and reproduce existing methods.
- Support multiple datasets. We support new datasets by giving a intermediate JSON format.
- Support feature-based training and end-to-end training. The feature-based training can easily be extended to end-to-end training with raw video input, and the video backbone can be easily replaced.
| Feature based | End to end |
|---|---|
| AvgPool | E2E-Spot |
| MaxPool | |
| NetVLAD | |
| NetRVLAD | |
| CALF | |
| AvgPool++ | |
| MaxPool++ | |
| NetVLAD++ | |
| NetRVLAD++ |
Please refer to install.md for installation and data preparation.
Please refer to usage.md for details of training and evaluation scripts.
Note: - We have added few documentation guides inside different parts of the library to make working with the library easier. Like this is an instruction for class intialization inside the configs directory. So please try to follow them if you get issues while working with the library. :)
All the things that need to be done in the future is in roadmap.md.
If you think this repo is helpful, please cite us:
@misc{benzakour2024oslactionspottingunifiedlibraryaction,
title={OSL-ActionSpotting: A Unified Library for Action Spotting in Sports Videos},
author={Yassine Benzakour and Bruno Cabado and Silvio Giancola and Anthony Cioppa and Bernard Ghanem and Marc Van Droogenbroeck},
year={2024},
eprint={2407.01265},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2407.01265},
}If you have any questions, please contact: [email protected].