Research map / 3D vision / Face and shape
3DCoMPaT ++ : An Improved Large-Scale 3D Vision Dataset for Compositional Recognition
Habib Slim, Xiang Li, Yuchen Li, Mahmoud Ahmed, Mohamed Ayman, Ujjwal Upadhyay, Ahmed Abdelreheem, Arpit Prajapati, Suhail Pothigara, Peter Wonka, Mohamed Elhoseiny
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 shape-category prior in training and inference | CurveNet · fine-grained 3D part segmentation | shape-agnostic mIOU: 53.09 -> 59.61 | Improved |
| add shape-category prior in training and inference | PCT · fine-grained 3D part segmentation | shape-agnostic mIOU: 49.09 -> 56.14 | Improved |
| add shape-category prior in training and inference | CurveNet · coarse-grained 3D part segmentation | shape-agnostic mIOU: 76.32 -> 83.85 | Improved |
| use coarse-grained instead of fine-grained part labels | CurveNet · 3D part segmentation (no shape prior) | shape-agnostic mIOU: 53.09 -> 76.32 | Improved |
| provide ground-truth shape category (shape-informed mIOU) instead of shape-agnostic | 3D part segmentation models · fine-grained 3D part segmentation | mIOU: 53.09 -> 81.37 | Improved |
| use ResNet-50 instead of ResNet-18 | 2D shape classification · 3DCoMPaT++ renders | top-1 accuracy: 76.27 -> 90.20 | Improved |
| use 2D/3D bidirectional-projection fusion (BPNet) instead of late fusion of separate 2D and 3D models | GCR baseline · fine-grained GCR | Grounded-value-all: 3.84 -> 15.41 | Improved |
| use RGB 6D point input with PointNet++ instead of BPNet | GCR · coarse-grained GCR | Grounded-value-all: 50.44 -> 58.99 | Improved |
| use RGB PointNet++ instead of BPNet | GCR · fine-grained GCR | Value-all: 27.74 -> 25.36 | Hurt |
| increase number of style compositions per shape from 1 to 5 and 20 (ablation) | BPNet · GCR compositional metrics | Value-all and grounded metrics | Improved |
| increase number of style compositions from 20 to 50 (ablation) | BPNet · GCR compositional metrics | Value-all and grounded metrics | No clear effect |
Closest papers by meaning
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.94)
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.87)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.87)
- Uni3D: Exploring Unified 3D Representation at Scale (ICLR 2024 · similarity 0.87)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.86)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 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.
