Research map / Label-efficient and robust learning / OOD and anomaly detection
GS-CLIP: Zero-shot 3D Anomaly Detection by Geometry-Aware Prompt and Synergistic View Representation Learning
Zehao Deng, An Liu, Yan 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 |
|---|---|---|---|
| fuse rendered and depth views with the Synergistic Refinement Module (ablation) | GS-CLIP · MVTec3D-AD | O-AUROC: 80.9 -> 82.3 | Improved |
| add shape prompt (ablation) | GS-CLIP + SRM · MVTec3D-AD | O-AP: 93.9 -> 94.8 | Improved |
| add defect prompt (GDDM) (ablation) | GS-CLIP + SRM · MVTec3D-AD | P-PRO: 84.8 -> 85.6 | Improved |
| combine shape and defect prompts (ablation) | GS-CLIP + SRM · MVTec3D-AD | O-AP: 94.4 -> 96.2 | Improved |
| add cross-view consistency loss (ablation) | GS-CLIP · MVTec3D-AD | O-AUROC: 83.1 -> 83.6 | Improved |
| use depth map input instead of rendered image (ablation) | GS-CLIP single view · MVTec3D-AD | O-AP: 91.7 -> 91.4 | No clear effect |
| increase GDDM outlier points k from 4 to 12 (ablation) | GS-CLIP · MVTec3D-AD | P-PRO: 85.5 -> 86.4 | Improved |
| increase prototypes l to 32 (ablation) | GS-CLIP · MVTec3D-AD | P-PRO | Improved |
| increase number of views up to about 9 (ablation) | GS-CLIP · MVTec3D-AD | anomaly detection | Improved |
| use GS-CLIP instead of PointAD (one-vs-rest) | PointAD · MVTec3D-AD | O-AUROC: 82.0 -> 83.6 | Improved |
| add RGB images via multimodal fusion | GS-CLIP · MVTec3D-AD | O-AUROC: 83.6 -> 88.2 | Improved |
| use GS-CLIP instead of PointAD in inference time | PointAD · MVTec3D-AD | time per image (s): 0.40 -> 0.51 | Hurt |
Closest papers by meaning
- PointAD: Comprehending 3D Anomalies from Points and Pixels for Zero-shot 3D Anomaly Detection (NeurIPS 2024 · similarity 0.92)
- Align3D-AD: Cross-Modal Feature Alignment and Dual-Prompt Learning for Zero-shot 3D Anomaly Detection (NeurIPS 2026 · similarity 0.88)
- AA-CLIP: Enhancing Zero-Shot Anomaly Detection via Anomaly-Aware CLIP (CVPR 2025 · similarity 0.87)
- Towards Scalable 3D Anomaly Detection and Localization: A Benchmark via 3D Anomaly Synthesis and A Self-Supervised Learning Network (CVPR 2024 · similarity 0.86)
- Looking 3D: Anomaly Detection with 2D-3D Alignment (CVPR 2024 · similarity 0.86)
- Aligning and Prompting Anything for Zero-Shot Generalized Anomaly Detection (AAAI 2025 · 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.
