Video Anomaly Detection by Estimating Likelihood of Representations

Abstract

Video anomaly detection is a challenging task not only because it involves solving many sub-tasks such as motion representation, object localization and action recognition, but also because it is commonly considered as an unsupervised learning problem that involves detecting outliers. Traditionally, solutions to this task have focused on the mapping between video frames and their low-dimensional features, while ignoring the spatial connections of those features. Recent solutions focus on analyzing these spatial connections by using hard clustering techniques, such as K-Means, or applying neural networks to map latent features to a general understanding, such as action attributes. In order to solve video anomaly in the latent feature space, we propose a deep probabilistic model to transfer this task into a density estimation problem where latent manifolds are generated by a deep denoising autoencoder and clustered by expectation maximization. Evaluations on several benchmarks datasets show the strengths of our model, achieving outstanding performance on challenging datasets.

Margin-Aware Intra-Class Novelty Identification for Medical Images

“…There have been a lot of research works that summarize state-of-the-art anomaly detection methods, [15][16][17][18][19][20] generally the methods aiming for anomalous image data detection can be divided into the following three categories: i. AutoEncoder-based methods: AutoEncoder 13 (AE) models can help to extract significant embedding features by reconstructing the original images unsupervised. Trained with ID data, the architectures learn the "normality" and should lead to large reconstruction error when working on OOD dataset.…”