This repository contains a Jupyter Notebook that demonstrates a weakly-Supervised anomaly detection model for video-level anomaly detection on the UCF-Crime dataset. This method is implemented in Python.

One of the most famouse large-scale dataset video anomaly detection dataset with video-level labels is UCF-crime dataset that contains 1,900 untrimmed real-world outdoor and indoor surveillance videos. The total length of the videos is 128 hours, which contains 13 classes of anomalous events including: 1. Abuse, 2. Arrest, 3. Arson, 4. Assault, 5. Burglary, 6. Explosion, 7. Fighting, 8. Road Accident, 9. Robbery, 10. Shooting, 11. Stealing, 12. Shoplifting, 13. Vandalism.