CAEVAD

This is an official PyTorch implementation for "A convolutional autoencoder approach for weakly-supervised anomaly video detection"

• State-of-the-art on ShanghaiTech Campus dataset

Installation

Clone the repository.

git clone https://github.com/duchieuphan2k1/weakly-supervised-anomaly-video-detection.git
cd weakly-supervised-anomaly-video-detection

Download the VideoSwin feature of the ShanghaiTech Campus dataset by this link: shanghaitech-video-swin.

Thanks to this repo for the extracted Video Swin Feature above.

Download our trained model by this link: best_proposed_model.

Usage

Testing

python main.py --test 1 --modelpath [path_to_trained_model] --datafolder [your_data_folder]

• [path_to_trained_model]: the absolute path to the trained model, which can download by the link above
• [your_data_folder]: the absolute path to the data folder, which can download by the link shanghaitech-video-swin above

Thanks to RTFM for the starter code.