AI Paper Atlas

Research map / Video / Video segmentation and tracking

A Causal Inference Look at Unsupervised Video Anomaly Detection

Xiangru Lin, Yuyang Chen, Guanbin Li, Yizhou Yu

AAAI 2022 · Video › Video segmentation and tracking · approach: Weakly supervised video anomaly detection · 42 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add de-confounded training to the strong baseline (ablation)ResNet-50 UVAD model · UCSD Ped1frame-level AUC: 70.7% -> 73.9%Improved
Add de-confounded training to the strong baseline (ablation)ResNet-50 UVAD model · UCSD Ped2frame-level AUC: 81.2% -> 97.9%Improved
Add counterfactual temporal context ensemble to de-confounded training (ablation)ResNet-50 UVAD model · UCSD Ped1frame-level AUC: 73.9% -> 84.9%Improved
Add counterfactual temporal context ensemble to de-confounded training (ablation)ResNet-50 UVAD model · UCSD Ped2frame-level AUC: 97.9% -> 99.4%Improved
Use counterfactual feature replacement with the sliding-window mean rather than CTCE variant 1 (ablation)Counterfactual temporal context ensemble · UCSD Ped1frame-level AUC: 76.8% -> 84.9%Improved
Use counterfactual feature replacement with the sliding-window mean rather than CTCE variant 2's zero feature (ablation)Counterfactual temporal context ensemble · UCSD Ped2frame-level AUC: 94.4% -> 99.4%Improved
Use focal loss instead of mean squared error loss (ablation)ResNet-50 UVAD model · UCSD Ped1frame-level AUC: 79.0% -> 84.9%Improved
Use focal loss instead of binary cross entropy loss (ablation)ResNet-50 UVAD model · UCSD Ped2frame-level AUC: 97.6% -> 99.4%Improved
Use I3D-RGB instead of ResNet-50 as the backbone (ablation)Ours + DCFD + CTCE · Avenueframe-level AUC: 87.3% -> 90.3%Improved
Use I3D-RGB instead of ResNet-50 as the backbone (ablation)Ours + DCFD + CTCE · UCSD Ped1frame-level AUC: 84.9% -> 84.9%No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy