Research map / Video / Video segmentation and tracking
Comprehensive Regularization in a Bi-directional Predictive Network for Video Anomaly Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| add backward prediction using backward-time information (ablation) | Bidirectional Predictive Networks · UCSD Ped2 | AUC: 97.3% -> 97.9% | Helped |
| add backward prediction using backward-time information (ablation) | Bidirectional Predictive Networks · CUHK Avenue | AUC: 89.6% -> 90.0% | Helped |
| add backward prediction using backward-time information (ablation) | Bidirectional Predictive Networks · Traffic-Train | AUC: 64.2% -> 66.6% | Helped |
| add a second forward predictor instead of a backward predictor (ablation) | Bidirectional Predictive Networks · UCSD Ped2 and CUHK Avenue | AUC: 97.3% -> 97.3%; 89.6% -> 89.5% | No clear effect |
| add sequence-wise discriminator for temporal consistency (ablation) | Bidirectional Predictive Networks · UCSD Ped2, CUHK Avenue, and Traffic-Train | AUC: 0.2%, 0.1% and 1.0% AUC gain | Helped |
| add multi-modal discriminator for association consistency (ablation) | Bidirectional Predictive Networks · UCSD Ped2, CUHK Avenue, and Traffic-Train | AUC: 0.2%, 0.2%, and 0.9% | Helped |
| add Laplacian pyramid loss to MSE (ablation) | Appearance prediction · UCSD Ped2, CUHK Avenue, and Traffic-Train | AUC: 98.0% -> 98.3%; 90.1% -> 90.3%; 67.1% -> 68.5% | Helped |
| use a 5-frame video event instead of 3, 7, or 9 frames (ablation) | Bidirectional Predictive Networks · UCSD Ped2, CUHK Avenue, and Traffic-Train | AUC: 93.5%, 97.1%, 96.6% -> 98.3%; 88.1%, 89.3%, 89.1% -> 90.3%; 65.7%, 66.7%, 64.5% -> 68.5% | Mixed |
| use the proposed method rather than VEC | Video anomaly detection · UCSD Ped2 | frame-level AUC: 97.3% -> 98.3% | Helped |
| use the proposed method rather than VEC | Video anomaly detection · CUHK Avenue | frame-level AUC: 89.6% -> 90.3% | Helped |
| use the proposed method rather than VEC | Video anomaly detection · ShanghaiTech | frame-level AUC: 74.8% -> 78.1% | Helped |
Closest papers by meaning
- Dual Conditioned Motion Diffusion for Pose-Based Video Anomaly Detection (AAAI 2025 · similarity 0.88)
- Multi-Scale Video Anomaly Detection by Multi-Grained Spatio-Temporal Representation Learning (CVPR 2024 · similarity 0.87)
- Just Dance with pi! A Poly-modal Inductor for Weakly-supervised Video Anomaly Detection (CVPR 2025 · similarity 0.87)
- Video Anomaly Detection with Motion and Appearance Guided Patch Diffusion Model (AAAI 2025 · similarity 0.87)
- Noise-Resistant Video Anomaly Detection via RGB Error-Guided Multiscale Predictive Coding and Dynamic Memory (CVPR 2025 · similarity 0.87)
- Track Any Anomalous Object:A Granular Video Anomaly Detection Pipeline (CVPR 2025 · similarity 0.86)
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.