Dynamic Distinction Learning: Adaptive Pseudo Anomalies for Video Anomaly Detection

This repository contains the implementation of the paper:

Dynamic Distinction Learning: Adaptive Pseudo Anomalies for Video Anomaly Detection

Authors: Demetris Lappas, Vasileios Argyriou, Dimitrios Makris

Affiliation: Kingston University, London, UK, School of Computer Science and Mathematics

Abstract

We introduce Dynamic Distinction Learning (DDL) for Video Anomaly Detection, a novel video anomaly detection methodology that combines pseudo-anomalies, dynamic anomaly weighting, and a distinction loss function to improve detection accuracy. By training on pseudo-anomalies, our approach adapts to the variability of normal and anomalous behaviors without fixed anomaly thresholds. Our model showcases superior performance on the Ped2, Avenue, and ShanghaiTech datasets, where individual models are tailored for each scene. These achievements highlight DDL’s effectiveness in advancing anomaly detection, offering a scalable and adaptable solution for video surveillance challenges. Our work can be found on: GitHub, Paper

Novelty Objects

The Pseudo Anomaly Creator and the Pseudo Loss (the linear combitnation of the reconstruction loss and distinction loss) can be found in ".\objects\pseudo_loss.py".

Installation

To install the necessary packages, you can use the provided requirements.txt file. Run the following command:

pip install -r requirements.txt

Tracked Bounding Boxes

The tracked_bounding_boxes.json files for Ped2 and Avenue can be downloaded here. Place the json files in their respective dataset training folders, and replace the "/path/to/training/frames" string within each json file with the path to your training folder.

Usage

Training

To train the model, use the train.py script. You can adjust the parameters as needed.

python train.py --dataset_name <DATASET_NAME> --training_batch_size <BATCH_SIZE> --temporal <TEMPORAL_LENGTH> --resize_height <HEIGHT> --resize_width <WIDTH> --epochs <EPOCHS> --lr <LEARNING_RATE> --device <DEVICE> --channels <CHANNELS>

Example:

python train.py --dataset_name ped2 --training_batch_size 10 --temporal 3 --resize_height 256 --resize_width 256 --epochs 50 --lr 0.01 --device cuda --channels 3

Testing

To test the model, use the test.py script. Make sure to provide the same parameters used during training.

python test.py --dataset_name <DATASET_NAME> --temporal <TEMPORAL_LENGTH> --resize_height <HEIGHT> --resize_width <WIDTH> --device <DEVICE> --channels <CHANNELS>

Example:

python test.py --dataset_name ped2 --temporal 3 --resize_height 256 --resize_width 256 --device cuda --channels 3

References

Please refer to the original paper for detailed explanations and experimental results:

Dynamic Distinction Learning: Adaptive Pseudo Anomalies for Video Anomaly Detection by Demetris Lappas, Vasileios Argyriou, Dimitrios Makris

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

We acknowledge Kingston University, London, UK, for their support and resources provided for this research.