Research map / Point clouds / Point cloud classification
PatchAlign3D: Local Feature Alignment for Dense 3D Shape Understanding
Souhail Hadgi, Bingchen Gong, Ramana Sundararaman, Emery Pierson, Lei Li, Peter Wonka, Maks Ovsjanikov
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 Stage 1 2D-to-3D distillation before text alignment (ablation) | PatchAlign3D · ShapeNetPart | mIoU: 50.5 -> 56.9 | Improved |
| add Stage 1 distillation (two-stage vs Stage 2 only) (ablation) | PatchAlign3D · ShapeNetPart | cIoU: 50.0 -> 53.1 | Improved |
| train both losses jointly instead of sequentially (ablation) | PatchAlign3D · ShapeNetPart | mIoU: 56.9 -> 50.2 | Hurt |
| joint training vs Stage 2 only (ablation) | PatchAlign3D · ShapeNetPart | cIoU: 50.0 -> 48.6 | Hurt |
| replace COPS (rendering) with PatchAlign3D feed-forward encoder | COPS · ShapeNetPart | mIoU: 25.6 -> 56.9 | Improved |
| replace Find3D with PatchAlign3D on same data | Find3D · ShapeNetPart | mIoU: 23.3 -> 56.9 | Improved |
| replace Find3D with PatchAlign3D | Find3D · FAUST | mIoU: 63.2 -> 67.8 | Improved |
| replace Find3D with PatchAlign3D | Find3D · FAUST (torso) | mIoU: 52.0 -> 49.7 | Hurt |
| replace Find3D with PatchAlign3D | Find3D · Objaverse-General unseen | mIoU: 34.6 -> 35.61 | Improved |
| replace Find3D with PatchAlign3D | Find3D · ScanObjectNN | mIoU: 18.8 -> 22.7 | Improved |
| replace PartSLIP with PatchAlign3D | PartSLIP · PartNetE | cIoU: 36.4 -> 42.2 | Improved |
| feed-forward single pass instead of rendering | COPS · inference time | seconds: 1.38 -> 0.4 | 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.
- PointCLIPv2 — PointCLIP V2: Prompting CLIP and GPT for Powerful 3D Open-world Learning (ICCV 2023) · better on 2 of 2 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
- 3D-PLOT-LLM: Part-Level Object Tokens for 3D Large Language Models (NeurIPS 2026 · similarity 0.91)
- More Text, Less Point: Towards 3D Data-Efficient Point-Language Understanding (AAAI 2025 · similarity 0.90)
- Tango3D: Towards Alignment for Global and Local 2D-3D Correspondence (NeurIPS 2026 · similarity 0.89)
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.89)
- Merge3D: Efficient 3D Multimodal LLMs via Joint 2D-3D Token Merging (CVPR 2026 · similarity 0.89)
- Part-X-MLLM: Part-aware 3D Multimodal Large Language Model (ICLR 2026 · similarity 0.89)
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.
