Research map / Video / Video segmentation and tracking
Learning Normal Dynamics in Videos With Meta Prototype Network
Check 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 DPU to AE (ablation) | AE · UCSD Ped1 | AUC (%): 83.2 -> 85.1 | Helped |
| Add DPU to AE (ablation) | AE · UCSD Ped2 | AUC (%): 95.1 -> 96.9 | Helped |
| Add DPU to AE (ablation) | AE · CUHK Avenue | AUC (%): 84.0 -> 89.5 | Helped |
| Add DPU to AE (ablation) | AE · ShanghaiTech | AUC (%): 66.7 -> 73.8 | Helped |
| Combine frame-prediction and feature-reconstruction anomaly scores (ablation) | DPU-based AE · UCSD Ped2 | AUC (%): 92.6 / 96.2 -> 96.9 | Helped |
| Use frame-prediction score alone rather than combine it with feature reconstruction (ablation) | DPU-based AE · UCSD Ped2 | AUC (%): 96.9 -> 92.6 | Hurt |
| Place DPU at full input resolution rather than lower resolution (ablation) | DPU-based AE · UCSD Ped2 feature-reconstruction score | AUC: 81.18, 84.66, 86.72 -> 89.19 | Helped |
| Increase prototype count from 1 to 10 (ablation) | DPU · UCSD Ped2 | AUC (Overall): 95.22 -> 96.90 | Helped |
| Increase prototype count from 10 to 20 or 40 (ablation) | DPU · UCSD Ped2 | AUC (Overall): 96.90 -> 96.03; 95.74 | Hurt |
| Use meta-trained MPU rather than the non-adapted model | DPU-based AE · UCSD Ped2, Shanghai Tech cross-dataset, 10-shot vs 0-shot | AUC (%): 90.17 -> 95.75 | Helped |
| Use meta-trained MPU rather than the non-adapted model | DPU-based AE · CUHK Avenue, UCF-Crime cross-dataset, 10-shot vs 0-shot | AUC (%): 82.26 -> 85.91 | Helped |
| Use the proposed MPU instead of rGAN (Meta) | Cross-dataset anomaly detection · UCSD Ped1, Shanghai Tech cross-dataset, 1-shot | AUC (%): 80.6 -> 78.54 | Hurt |
Closest papers by meaning
- Self-Distilled Masked Auto-Encoders are Efficient Video Anomaly Detectors (CVPR 2024 · similarity 0.87)
- Noise-Resistant Video Anomaly Detection via RGB Error-Guided Multiscale Predictive Coding and Dynamic Memory (CVPR 2025 · similarity 0.86)
- Dual Conditioned Motion Diffusion for Pose-Based Video Anomaly Detection (AAAI 2025 · similarity 0.84)
- Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularity (CVPR 2025 · similarity 0.84)
- Video Anomaly Detection with Motion and Appearance Guided Patch Diffusion Model (AAAI 2025 · similarity 0.83)
- UCF-Crime-DVS: A Novel Event-Based Dataset for Video Anomaly Detection with Spiking Neural Networks (AAAI 2025 · similarity 0.83)
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.