Research map / Video / Video segmentation and tracking
Video Anomaly Detection via Sequentially Learning Multiple Pretext Tasks
Chenrui Shi, Che Sun, Yuwei Wu, Yunde Jia
Read the paperCheck your idea against 137,217 papers
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 changed | Where | Measured | Effect |
|---|---|---|---|
| learn pretext tasks sequentially (Pre-Rec-Cls) instead of simultaneously with equivalent weights 3:2:1 (ablation) | video anomaly detection model · Ped2 | frame-level AUC (%): 87.97 -> 97.57 | Improved |
| learn pretext tasks sequentially (Pre-Rec-Cls) instead of simultaneously (ablation) | video anomaly detection model · Avenue | frame-level AUC (%): 80.21 -> 90.85 | Improved |
| order tasks easy-to-hard (Pre-Rec-Cls) instead of Rec-Cls-Pre (ablation) | sequential learning · Ped2 | frame-level AUC (%): 92.02 -> 97.57 | Improved |
| order tasks easy-to-hard (Pre-Rec-Cls) instead of Cls-Rec-Pre (ablation) | sequential learning · Avenue | frame-level AUC (%): 75.51 -> 90.85 | Improved |
| add more pretext losses simultaneously beyond prediction loss (ablation) | memory auto-encoder · Ped2 | frame-level AUC (%): 92.29 -> 83.85 | Hurt |
| add more pretext losses simultaneously beyond prediction loss (ablation) | memory auto-encoder · Avenue | frame-level AUC (%): 79.15 -> 76.51 | Hurt |
| compare method to AD-Con on Ped2 | Ours · UCSD Ped2 (RGB only) | frame-level AUC (%): 98.1 -> 97.6 | Hurt |
| compare method to AD-Con on Avenue and ShanghaiTech | Ours · Avenue | frame-level AUC (%): 88.8 -> 90.9 | Improved |
Closest papers by meaning
- Anomaly Detection in Video via Self-Supervised and Multi-Task Learning (CVPR 2021 · similarity 0.90)
- Learning Memory-Guided Normality for Anomaly Detection (CVPR 2020 · similarity 0.89)
- A Discriminative Framework for Anomaly Detection in Large Videos (ECCV 2016 · similarity 0.88)
- Learning Anomalies with Normality Prior for Unsupervised Video Anomaly Detection (ECCV 2024 · similarity 0.88)
- Prompt-Enhanced Multiple Instance Learning for Weakly Supervised Video Anomaly Detection (CVPR 2024 · similarity 0.86)
- Towards Multi-Domain Learning for Generalizable Video Anomaly Detection (NeurIPS 2024 · 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.
