REU Week 3: Real-Time Video Anomaly Detection Project Leader:Praveen Tirupattur Urvi Gianchandani
Datasets Papers Experiment Ideas Next Steps Overview
Datasets UCSD Ped1 & Ped2 CUHK Avenue Subway (Entrance & Exit) ShanghaiTech Campus
Abnormal Event Detection at 150 FPS in MATLAB Cewu Lu, Jianping Shi, Jiaya Jia (1) Sparse combination learning: sparsity coded directly as combination of basis vectors Training: learn dictionary by adding sparse combinations Testing: use reconstruction error to detect anomalies
Weixin Luo, Wen Liu, Shenghua Gao (2) A Revisit of Sparse Coding Based Anomaly Detection in Stacked RNN Framework Weixin Luo, Wen Liu, Shenghua Gao (2) Temporally coherent sparse coding: similar neighboring frames will have similar sparse coefficients
Abnormal Event Detection in Videos using Spatiotemporal Autoencoder Yong Shean Chong, Yong Haur Tay (3) Spatial encoder → temporal encoder-decoder (3 ConvLSTMs) → spatial decoder Spatial: extract features Temporal: learn temporal features of encoded spatial structure
AUC Comparisons Ped1 Ped2 Avenue Cewu Lu (1) 91.8 -- Weixin Luo (2) 92.2 81.7 Yong Shean Chong (3) 89.9 87.4 80.3
Experiment: Simeple Autoencoder
Ideas Learn normal feature of specific object within a scene - Object-centric Auto-encoders and Dummy Anomalies for Abnormal Event Detection in Video (4) Use optical flow(reconstruction) and patch-based generator to localize spatial features
Running experiments on datasets Exploring optical flow Next Steps
References http://shijianping.me/abnormal_iccv13.pdf http://openaccess.thecvf.com/content_ICCV_2017/papers/Luo_A_Revisit_of_ICCV_2017_paper.pdf https://arxiv.org/pdf/1701.01546.pdf https://arxiv.org/pdf/1812.04960.pdf