Research map / Detection and segmentation / Open-set and few-shot detection
LayoutAD: Exploring Semantic-Geometric Misalignment Reasoning for Scene Layout Anomaly Detection
Zhichao Zeng, Jiasheng Zhang, Jiyun Sun, Jiangtao Cui, Xiaotian Qiao
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 both semantic and geometric features (vs geometric only) (ablation) | LayoutAD · COCOAD | I-AUROC: 0.473 -> 0.586 | Improved |
| use both semantic and geometric features (vs semantic only) (ablation) | LayoutAD · COCOAD | I-AUROC: 0.535 -> 0.586 | Improved |
| add cross-graph transformer (ablation) | LayoutAD · COCOAD | A-P-AUROC: 0.815 -> 0.883 | Improved |
| add cross-graph transformer (ablation) | LayoutAD · COCOAD | I-AUROC: 0.509 -> 0.586 | Improved |
| add GNN message passing (ablation) | LayoutAD · COCOAD | I-AUROC: 0.546 -> 0.586 | Improved |
| add global conditional modeling (ablation) | LayoutAD · COCOAD | I-AUROC: 0.522 -> 0.586 | Improved |
| add attribute-level likelihood loss (ablation) | LayoutAD · COCOAD | I-AUROC: 0.527 -> 0.586 | Improved |
| add relation-level likelihood loss (ablation) | LayoutAD · COCOAD | I-AUROC: 0.530 -> 0.586 | Improved |
| add cross-graph transformer (ablation) | LayoutAD · COCOAD | P-AUROC: 0.794 -> 0.871 | Improved |
| use LayoutAD object-level reasoning instead of motion-based detectors, no temporal modeling | video anomaly detection · StreetScene | RBDC (%): 24.3% -> 25.4% vs EVAL | Improved |
| use structural anomaly detectors (PatchCore, UniAD, etc.) instead of layout reasoning | scene layout anomaly detection · COCOAD | I-AUROC: 0.479 to 0.573 vs 0.586 | Hurt |
Closest papers by meaning
- Looking 3D: Anomaly Detection with 2D-3D Alignment (CVPR 2024 · similarity 0.87)
- Towards Generalizable 3D Anomaly Detection via Relational Inconsistency Modeling (NeurIPS 2026 · similarity 0.87)
- Odd-One-Out: Anomaly Detection by Comparing with Neighbors (CVPR 2025 · similarity 0.86)
- UniVAD: A Training-free Unified Model for Few-shot Visual Anomaly Detection (CVPR 2025 · similarity 0.85)
- Align3D-AD: Cross-Modal Feature Alignment and Dual-Prompt Learning for Zero-shot 3D Anomaly Detection (NeurIPS 2026 · similarity 0.85)
- Few Shot Part Segmentation Reveals Compositional Logic for Industrial Anomaly Detection (AAAI 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.
