Research map / Label-efficient and robust learning / OOD and anomaly detection
Few-Shot Scene-Adaptive Anomaly Detection
Yiwei Lu, Frank Yu, Mahesh Kumar Krishna Reddy, Yang Wang
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 |
|---|---|---|---|
| add MAML-based scene adaptation instead of fine-tuning | r-GAN · Shanghai Tech 7 scenes, K=10 | AUC (%): 71.59 -> 77.36 | Improved |
| add MAML-based scene adaptation instead of fine-tuning | r-GAN · Shanghai Tech 7 scenes, K=1 | AUC (%): 71.61 -> 74.51 | Improved |
| add MAML meta-learning rather than fine-tuning, cross-dataset (UCF crime meta-training) | r-GAN · UCSD Ped2, K=1 | AUC (%): 65.58 -> 83.08 | Improved |
| add MAML adaptation rather than fine-tuning, cross-dataset | r-GAN · UR Fall depth images, K=10 (Shanghai Tech meta-training) | AUC (%): 62.89 -> 83.24 | Improved |
| adapt to the target scene with K frames versus no adaptation (fine-tuning) | r-GAN pre-trained · UR Fall, K=10 (Shanghai Tech meta-training) | AUC (%): 64.08 -> 62.89 | No clear effect |
| sample 5 tasks per epoch instead of 1 during meta-training (ablation) | MAML scene-adaptive · UCSD Ped1, K=1 | AUC (%): 79.94 -> 80.6 | Improved |
| sample 5 tasks per epoch instead of 1 (ablation) | MAML scene-adaptive · CUHK Avenue, K=10 | AUC (%): 77.31 -> 78.79 | Improved |
| sample 1 task per epoch instead of fine-tuning (ablation) | MAML scene-adaptive · UCSD Ped2, K=1 | AUC (%): 85.64 -> 90.73 | Improved |
| use the GAN-based backbone r-GAN instead of r-VAE | backbone · Shanghai Tech (standard) | AUC (%): 72.7 -> 77.9 | Improved |
| use more shots K from 1 to 10 for fine-tuning baseline | Fine-tuned baseline · Shanghai Tech 7 scenes | AUC (%): 71.61 -> 71.59 | No clear effect |
Closest papers by meaning
- Future Frame Prediction Network for Video Anomaly Detection (TPAMI 2021 · similarity 0.90)
- Video Anomaly Detection with Motion and Appearance Guided Patch Diffusion Model (AAAI 2025 · similarity 0.87)
- One-to-Normal: Anomaly Personalization for Few-shot Anomaly Detection (NeurIPS 2024 · similarity 0.87)
- Track Any Anomalous Object:A Granular Video Anomaly Detection Pipeline (CVPR 2025 · similarity 0.86)
- MetaUAS: Universal Anomaly Segmentation with One-Prompt Meta-Learning (NeurIPS 2024 · similarity 0.86)
- AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 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.
