Research map / Video / Video segmentation and tracking
Video Anomaly Detection by Solving Decoupled Spatio-Temporal Jigsaw Puzzles
Guodong Wang, Yunhong Wang, Jie Qin, Dongming Zhang, Xiuguo Bao, Di Huang
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 |
|---|---|---|---|
| use multi-label position prediction instead of multi-class permutation classification (same 5040/5040 permutations) (ablation) | jigsaw VAD · ShanghaiTech Campus | AUROC %: 78.5 -> 82.0 | Improved |
| use multi-label instead of multi-class (same 5040/5040 permutations) (ablation) | jigsaw VAD · CUHK Avenue | AUROC %: 87.5 -> 89.5 | Improved |
| increase permutations from 504/504 to 5040/5040 (ablation) | multi-label jigsaw · ShanghaiTech Campus | AUROC %: 79.7 -> 82.0 | Improved |
| use full spatial permutations (362880) instead of 5040 (ablation) | multi-label jigsaw · CUHK Avenue | AUROC %: 89.5 -> 92.2 | Improved |
| use full spatial permutations (362880) instead of 5040 (ablation) | multi-label jigsaw · ShanghaiTech Campus | AUROC %: 82.0 -> 83.2 | Improved |
| increase patches from n^2=4 to n^2=9 (l=7) (ablation) | jigsaw VAD · ShanghaiTech Campus | AUROC %: 80.2 -> 83.2 | Improved |
| increase patches from n^2=9 to n^2=16 (l=9) (ablation) | jigsaw VAD · ShanghaiTech Campus | AUROC %: 84.3 -> 80.4 | Hurt |
| increase frames from l=7 to l=9 (n^2=9) (ablation) | jigsaw VAD · ShanghaiTech Campus | AUROC %: 83.2 -> 84.3 | Improved |
| increase frames from l=7 to l=9 (n^2=9) (ablation) | jigsaw VAD · CUHK Avenue | AUROC %: 92.2 -> 89.2 | Hurt |
| add spatial puzzle to temporal-only training (ablation) | jigsaw VAD · ShanghaiTech Campus | AUROC %: 78.6 -> 82.7 | Improved |
| test with both temporal and spatial solvers instead of temporal only (ablation) | jigsaw VAD · ShanghaiTech Campus | AUROC %: 82.7 -> 84.3 | Improved |
| replace rotation/arrow-of-time pretexts with spatio-temporal jigsaw (ablation) | pretext task · ShanghaiTech Campus | AUROC %: 72.9 -> 84.3 | Improved |
Closest papers by meaning
- Multi-Scale Video Anomaly Detection by Multi-Grained Spatio-Temporal Representation Learning (CVPR 2024 · similarity 0.88)
- Track Any Anomalous Object:A Granular Video Anomaly Detection Pipeline (CVPR 2025 · similarity 0.86)
- Advancing Video Anomaly Detection: A Concise Review and a New Dataset (NeurIPS 2024 · similarity 0.85)
- VADTree: Explainable Training-Free Video Anomaly Detection via Hierarchical Granularity-Aware Tree (NeurIPS 2025 · similarity 0.85)
- Kernel-Aware Graph Prompt Learning for Few-Shot Anomaly Detection (AAAI 2025 · similarity 0.85)
- Solving Masked Jigsaw Puzzles with Diffusion Vision Transformers (CVPR 2024 · similarity 0.85)
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.
