Research map / Point clouds / Shape and surface completion
TransCAD: A Hierarchical Transformer for CAD Sequence Inference from Point Clouds
Elona Dupont, Kseniya Cherenkova, Dimitrios Mallis, Gleb A Gusev, Anis Kacem, Djamila Aouada
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 hierarchical loop-extrusion decoding (ablation) | TransCAD · DeepCAD test | APCS: 0.687 -> 0.708 (w/o refining) / 0.732 (full) | Improved |
| add loop refiner (ablation) | TransCAD · DeepCAD test | APCS: 0.708 -> 0.732 | Improved |
| add loop refiner (also on extrusion APCS components) (ablation) | TransCAD · DeepCAD test | APCS Ext. component: 0.755 -> 0.818 | Improved |
| TransCAD vs DeepCAD under Perlin noise | point cloud noise · DeepCAD (Perlin noise) | APCS: 0.550 -> 0.604 | Improved |
| increase Perlin noise magnitude by 50% | TransCAD · DeepCAD (Perlin noise) | APCS: 0.604 -> 0.532 | Hurt |
| finetune 2 epochs on noisy training data | TransCAD · DeepCAD (Perlin noise) | APCS: 0.604 -> 0.724 | Improved |
| reduce input points from 4096 to 256 | TransCAD · DeepCAD test | APCS: 0.732 -> 0.673 | Hurt |
| use ground-truth loop/extrusion types at inference (ablation) | TransCAD · DeepCAD test | APCS: 0.732 -> 0.790 | Improved |
| TransCAD vs retrieval baseline on chamfer distance | TransCAD · DeepCAD test | Chamfer distance (x10^3): 2.8 (retrieval) vs 4.5 | Hurt |
| TransCAD vs DeepCAD on Fusion360 APCS | TransCAD · Fusion360 (cross-dataset) | APCS: 0.360 -> 0.365 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports TransCAD as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- ReACT: Reward-informed Autoregressive Decision CAD Transformer (AAAI 2026 · better on 5 of 5 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
- Point2CAD: Reverse Engineering CAD Models from 3D Point Clouds (CVPR 2024 · similarity 0.90)
- CAD-SIGNet: CAD Language Inference from Point Clouds using Layer-wise Sketch Instance Guided Attention (CVPR 2024 · similarity 0.90)
- MiCADangelo: Fine-Grained Reconstruction of Constrained CAD Models from 3D Scans (NeurIPS 2025 · similarity 0.88)
- Draw Step by Step: Reconstructing CAD Construction Sequences from Point Clouds via Multimodal Diffusion. (CVPR 2024 · similarity 0.87)
- CADCrafter: Generating Computer-Aided Design Models from Unconstrained Images (CVPR 2025 · similarity 0.87)
- From 2D CAD Drawings to 3D Parametric Models: A Vision-Language Approach (AAAI 2025 · similarity 0.87)
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.
