Year: 2025
Type: conference-paper
Abstract: Video anomaly detection aims to develop automated models capable of identifying abnormal events in surveillance videos. The benchmark setup for this task is extremely challenging due to: i) the limited size of the training sets, ii) weak supervision provided in terms of video-level labels, and iii) intrinsic class imbalance induced by the scarcity of abnormal events. In this work, we show that distilling knowledge from aggregated representations of multiple backbones into a single-backbone Student model achieves state-of-the-art performance. In particular, we develop a bi-level distillation approach along with a novel disen-tangled cross-attention-based feature aggregation network. Our proposed approach, DAKD (Distilling Aggregated Knowledge with Disentangled Attention), demonstrates superior performance compared to existing methods across multiple benchmark datasets. Notably, we achieve significant improvements of 1.36%, 0.78%, and 7.02% on the UCF-Crime, ShanghaiTech, and XD-Violence datasets, respectively.
Authors: Jash Dalvi, Ali Dabouei, Gunjan Dhanuka, Min Xu
Institution: Carnegie Mellon University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: National Science Foundation
Awards: Tools4Cells: Machine-learning aided morphodynamics characterization of stem cell differentiation using label-free microscopies, Collaborative Research: III: Medium: Systematic De Novo Identification of Macromolecular Complexes in Cryo-Electron Tomography Images, III: Small: Improving automation and speed of macromolecule recognition and localization in cryo-electron tomography using unsupervised deep learning, IIBR Informatics: Reducing the training data annotation cost for learning-based macromolecule identification in cellular electron cryo-tomography, CAREER: Cryo-electron tomography derived multiscale integrative modeling of subcellular organization
OpenAlex ID: W4409262023