Research map / Detection and segmentation / Open-set and few-shot detection
Wavelet-Driven 3D Anomaly Detection under Pose-Agnostic and Sparse-View
Mingwen Shao, Qiao Zhang, Xinyuan Chen, Xiang Lv, Lingzhuang Meng, Chang Liu, Qinglin Zhan, Ling Jian
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 SWGM (structure-aware wavelet-optimized Gaussian modeling) (ablation) | Wave-Pose3D · PIAD-synt (20% views) | image AUROC: 80.05 -> 82.34 | Improved |
| add WPE (wavelet-based pose estimator) (ablation) | Wave-Pose3D · PIAD-synt (20% views) | image AUROC: 78.65 -> 82.34 | Improved |
| add WDAD (wavelet difference-aware anomaly detector) (ablation) | Wave-Pose3D · PIAD-synt (20% views) | image AUROC: 80.20 -> 82.34 | Improved |
| use SRFilter instead of Dropout (ablation) | SWGM · PIAD-synt (20% views) | image AUROC: 78.76 -> 82.34 | Improved |
| add SRFilter structure-aware region filtering (ablation) | SWGM · PIAD-synt (20% views) | image AUROC: 80.12 -> 82.34 | Improved |
| add high-frequency consistency loss (ablation) | SWGM · PIAD-synt (20% views) | image AUROC: 80.08 -> 82.34 | Improved |
| use wavelet matching verification instead of spatial matching (ablation) | WPE · PIAD-synt (20% views) | image AUROC: 80.80 -> 82.34 | Improved |
| use wavelet pose optimization instead of spatial optimization (ablation) | WPE · PIAD-synt (20% views) | image AUROC: 80.14 -> 82.34 | Improved |
| set SWGM balance coefficient omega to 0.6 versus 0.5 or 0.7 (ablation) | SWGM · PIAD-synt (20% views) | image AUROC: 81.89 / 82.34 / 82.21 (0.5 / 0.6 / 0.7) | Mixed |
| set WPE eta to 0.8 versus 0.7 or 0.9 (ablation) | WPE · PIAD-synt (20% views) | image AUROC: 82.01 / 82.34 / 81.89 (0.7 / 0.8 / 0.9) | Mixed |
| increase high-frequency weight beta in WDAD (alpha=1) (ablation) | WDAD · PIAD-synt (20% views) | image AUROC: 79.53 (beta 0.5), 80.36 (1), 81.33 (2), 82.34 (3), 82.28 (4) | Mixed |
| Wave-Pose3D vs DropGaussian with PIAD modules | Wave-Pose3D · PIAD-synt (20% views) | image AUROC: 77.8 -> 82.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PIAD — PIAD: Pose and Illumination agnostic Anomaly Detection (CVPR 2025) · better on 9 of 9 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · 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 0.87)
- A Unified Latent Schrodinger Bridge Diffusion Model for Unsupervised Anomaly Detection and Localization (CVPR 2025 · similarity 0.86)
- PointAD: Comprehending 3D Anomalies from Points and Pixels for Zero-shot 3D Anomaly Detection (NeurIPS 2024 · similarity 0.86)
- Registration is a Powerful Rotation-Invariance Learner for 3D Anomaly Detection (NeurIPS 2025 · similarity 0.85)
- PO3AD: Predicting Point Offsets toward Better 3D Point Cloud Anomaly Detection (CVPR 2025 · 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.
