Anomaly detection in videos

This repository contains source code for our AI course project about frame-level anomaly detection in videos.

Files

• oneclasssvm.ipynb : contains low level method based on HOG and One Class SVM.
• utilities.py : contains all the utilities for our version of the Object-centric autoencoder method.
• networks.py : contains the implementations of all the auto-encoders.
• train_model_v1.py : contains the training script of the auto-encoders.
• detect_anomalies_v1.py : contains the testing script (inference mode) of the Object-centric autoencoder method.

Requirements

• retinanet
• matplotlib 3.0.2
• seaborn 0.9.0
• opencv 3.4.2
• tensorflow 1.12.0
• scikit-image 0.14.2
• scikit-learn 0.20.3
• pandas 0.24.2

We installed retinanet from source and the rest of the packages with conda.

Contact

• [email protected]
• [email protected]
• [email protected]
• [email protected]
• [email protected]