This is an official PyTorch Implementation of "CLIP-Driven Multi-Scale Instance Learning for Weakly Supervised Video Anomaly Detection" in ICME 2024.
Please download the extracted I3D features for XD-Violence and UCF-Crime dataset from links below:
After downloading, please put it in the data/ folder (No need to unzip)
# For XD-Violence
python train.py --zip_feats data/xdviolence_i3d_w16_s16.zip --amp
# For UCF-Crime
python train.py --zip_feats data/ucfcrime_i3d_roc_ng_w16_s16.zip --amp@inproceedings{vad_cmsil,
author = {Qian, Zhangbin and Tan, Jiawei and Ou, Zhilong and Wang, Hongxing},
title = {CLIP-Driven Multi-Scale Instance Learning for Weakly Supervised Video Anomaly Detection},
booktitle = {2024 IEEE International Conference on Multimedia and Expo (ICME)},
pages = {1--6},
year = {2024}
}