Litcius/Paper detail

Learning Task-Specific Representation for Video Anomaly Detection with Spatial-Temporal Attention

Yang Liu, Jing Liu, Xiaoguang Zhu, Donglai Wei, Xiaohong Huang, Liang Song

2022ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)44 citationsDOI

Abstract

The automatic detection of abnormal events in surveillance videos with weak supervision has been formulated as a multiple instance learning task, which aims to localize the clips containing abnormal events temporally with the video-level labels. However, most existing methods rely on the features extracted by the pre-trained action recognition models, which are not discriminative enough for video anomaly detection. In this work, we propose a spatial-temporal attention mechanism to learn inter- and intra-correlations of video clips, and the boosted features are encouraged to be task-specific via the mutual cosine embedding loss. Experimental results on standard benchmarks demonstrate the effectiveness of the spatial-temporal attention, and our method achieves superior performance to the state-of-the-art methods.

Topics & Concepts

Computer scienceDiscriminative modelArtificial intelligenceTask (project management)Anomaly detectionEmbeddingPattern recognition (psychology)CLIPSRepresentation (politics)Feature learningTask analysisComputer visionMachine learningManagementLawEconomicsPolitical sciencePoliticsAnomaly Detection Techniques and ApplicationsHuman Pose and Action RecognitionVideo Surveillance and Tracking Methods
Learning Task-Specific Representation for Video Anomaly Detection with Spatial-Temporal Attention | Litcius