UWS4VAD is an attempt to unify common pratices in VAD methods, with support for both UCFC and XDV datasets, configured trough hydra, in a modular and experimental pipeline, setting ground for a centralised experimental playground and benchmark. Includes feature extraction for both visual (trough timm models) and audio (trough HEAR-based models).~
Important
Looking for contributions/suggestions of any kind. If you have interest in the project, please dont hesitate to contact, will be more than grateful for such.
conda env create -f environment.yml && conda activate uws4vadBasic overview of configuration setup
python main.py --helpRefer to wiki/config for a broader view of both configuration and usage.
Refer to wiki/Dev for instructions on how to contribute or help with development, as well as an updated to-do list.
Gratzie to author's works that are either part of this project, served as inspiration or contributed to VAD/VAU.
Refer to wiki/methods for a complete and updated list.
Give a shout if used <3
@misc{uws4vad,
author = {Zuble Barbas},
title = {A Unified WorkStation for Video Anomaly Detection.},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/zuble/uws4vad}},
year = {2024},
}