Dilip Mathew Thomas

This project reviewed different techniques for detecting anomalies in videos using unsupervised machine learning techniques. We explored the use of reconstruction models, predictive models, and deep-learning-based generative models for this project.

These reconstruction-based models build representations that minimize the reconstruction error of training samples from the normal distribution. Spatiotemporal predictive models take into account the spatiotemporal correlation by viewing videos as a spatiotemporal time series and learn representations that minimize the prediction error on spatiotemporal sequences. The generative models learn to generate samples from the training distribution while minimizing the reconstruction error as well as the distance between generated and training distribution. Each of these methods focuses on learning certain prior information that is useful for constructing the representation for the video anomaly detection task.