Research map / 3D vision / Face and shape
Patch2CAD: Patchwise Embedding Learning for In-the-Wild Shape Retrieval From a Single Image
Weicheng Kuo, Anelia Angelova, Tsung-Yi Lin, Angela Dai
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 |
|---|---|---|---|
| replace whole-shape embedding with patch-wise embedding (Patch2CAD vs Mask2CAD) | Mask2CAD · ScanNet 25K | mesh AP: 8.4 -> 10.3 | Improved |
| replace whole-shape embedding with patch-wise embedding, ground-truth 2D detections | Mask2CAD · ScanNet 25K (ground-truth 2D detections) | mesh AP: 10.5 -> 12.9 | Improved |
| patch-wise embedding evaluated on shape only (F-score) | Mask2CAD · ScanNet 25K (GT detections and pose, shape only) | mean F-score: 60.6 -> 63.8 | Improved |
| patch-wise embedding on Pix3D with exact CAD matches | Mask2CAD · Pix3D S1 | mesh AP: 33.2 -> 30.9 | Hurt |
| patch-wise embedding vs generative Mesh R-CNN on Pix3D | Mesh R-CNN · Pix3D S1 | mesh AP: 17.2 -> 30.9 | Improved |
| use patches of 1.0 (full object) with normals instead of Mask2CAD (ablation) | Mask2CAD · ScanNet 25K (patch size ablation) | mesh AP: 8.4 -> 9.4 | Improved |
| reduce patch size from 1.0 to 0.33 of ROI (ablation) | Patch2CAD · ScanNet 25K (patch size ablation) | mesh AP: 9.4 -> 10.3 | Improved |
| reduce patch size from 0.33 to 0.25 (ablation) | Patch2CAD · ScanNet 25K (patch size ablation) | mesh AP: 10.3 -> 10.0 | Hurt |
| increase query patches Kq from 1 to 9 (ablation) | Patch2CAD · ScanNet 25K (test-time query patches Kq) | mesh AP: 9.2 -> 10.3 | Improved |
| increase Kq from 9 to 12 (ablation) | Patch2CAD · ScanNet 25K (test-time query patches Kq) | mesh AP: 10.3 -> 10.2 | No clear effect |
| increase retrieved patches Kr from 1 to 48 (ablation) | Patch2CAD · ScanNet 25K (test-time retrieved patches Kr) | mesh AP: 9.3 -> 10.6 | Improved |
| increase Kr from 48 to 96 (ablation) | Patch2CAD · ScanNet 25K (test-time retrieved patches Kr) | mesh AP: 10.6 -> 10.6 | No clear effect |
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.
- Mask2CAD — Mask2CAD: 3D Shape Prediction by Learning to Segment and Retrieve (ECCV 2020) · better on 7 of 10 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
- Mask2CAD: 3D Shape Prediction by Learning to Segment and Retrieve (ECCV 2020 · similarity 0.90)
- PatchComplete: Learning Multi-Resolution Patch Priors for 3D Shape Completion on Unseen Categories (NeurIPS 2022 · similarity 0.89)
- FastCAD: Real-Time CAD Retrieval and Alignment from Scans and Videos (ECCV 2024 · similarity 0.89)
- Joint Embedding of 3D Scan and CAD Objects (ICCV 2019 · similarity 0.89)
- PatchNets: Patch-Based Generalizable Deep Implicit 3D Shape Representations (ECCV 2020 · similarity 0.88)
- KP-RED: Exploiting Semantic Keypoints for Joint 3D Shape Retrieval and Deformation (CVPR 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.
