Awesome

Codes_MPN

Official codes of CVPR21 paper: Learning Normal Dynamics in Videos with Meta Prototype Network (https://arxiv.org/abs/2104.06689)

MPN Framework

Paper Results on Unsupervised VAD

Paper Results on Few-shot VAD

Preparation

Please download the corresponding benchmarks in 'data' directory. Then prepare the environment as in requirement.txt. We have uploaded several trained models on online (Baidunetdisk(link：https://pan.baidu.com/s/1qcGmdmZlmAgqsAzw_i5BhA code：mapz) or Drive (https://drive.google.com/drive/folders/1ketomxctszHo7jpGQS3RxbZGq7e_M3e4?usp=sharing)).

Unsupervised Anomaly Detection Model Training

Run 'python Train.py' to train a model with DPU model.

Unsupervised Anomaly Detection Model Testing

Run 'python Test.py' to train a model with DPU model.

Meta-learning Anomaly Detection Model Training

Run 'python Train_meta.py' to train a model with MPU model.

Meta-learning Anomaly Detection Model Testing

Run 'python Test_meta.py' to test a model with MPU model.

If you find this work helpful, please cite:

@inproceedings{Lv2021MPN,
  author    = {Hui LV and
	       Chen Chen and
               Zhen Cui and
               Chunyan Xu and
               Yong Li and
               Jian Yang},
  title     = {Learning Normal Dynamics in Videos with Meta Prototype Network},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year      = {2021}
}