Research map / Detection and segmentation / Open-set and few-shot detection
PAD: A Dataset and Benchmark for Pose-agnostic 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 |
|---|---|---|---|
| reduce training samples from 100 to 70 views (ablation) | OmniposeAD · MAD-Real Gorilla | image-level AUROC: 93.6 -> 85.4 | Hurt |
| reduce training samples from 100 to 50 views (ablation) | OmniposeAD · MAD-Real Turtle | image-level AUROC: 99.1 -> 83.4 | Hurt |
| reduce training samples from 100 to 50 views (ablation) | OmniposeAD · MAD-Real Turtle | pixel-level AUROC: 95.4 -> 90.4 | Hurt |
| reduce training samples from 100 to 50 views (ablation) | OmniposeAD · MAD-Real Bear | image-level AUROC: 97.8 -> 87.8 | Hurt |
| reduce training samples from 100 to 50 views (ablation) | OmniposeAD · MAD-Real Bear | pixel-level AUROC: 99.3 -> 97.9 | Hurt |
| reduce training samples from 100 to 50 views (ablation) | OmniposeAD · MAD-Real Owl | image-level AUROC: 89.1 -> 89.1 | No clear effect |
| sparse training views, pixel versus image level (ablation) | OmniposeAD · MAD-Real | pixel vs image AUROC | Mixed |
| add ImageNet pretraining to baselines | OmniposeAD · MAD | AUROC | Helped |
| compare OmniposeAD with CFlow | OmniposeAD · MAD | pixel/image AUROC: +7.0/+19.6 | Helped |
| compare OmniposeAD with UniAD | OmniposeAD · MAD | pixel/image AUROC: +8.7/+28.7 | Helped |
| compare with methods not using additional training data | OmniposeAD · MAD | pixel/image AUROC: +29.2/+8.3 | Helped |
| pseudo-anomaly augmentation (DRAEM, Cutpaste) in PAD setting | OmniposeAD · MAD | AUROC: 58.0 | Hurt |
Closest papers by meaning
- PIAD: Pose and Illumination agnostic Anomaly Detection (CVPR 2025 · similarity 0.89)
- Odd-One-Out: Anomaly Detection by Comparing with Neighbors (CVPR 2025 · similarity 0.85)
- UniVAD: A Training-free Unified Model for Few-shot Visual Anomaly Detection (CVPR 2025 · similarity 0.85)
- PO3AD: Predicting Point Offsets toward Better 3D Point Cloud Anomaly Detection (CVPR 2025 · similarity 0.84)
- Advancing Video Anomaly Detection: A Concise Review and a New Dataset (NeurIPS 2024 · similarity 0.84)
- Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection (CVPR 2025 · similarity 0.84)
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.