Research map / 3D vision / Face and shape
Discovering 3D Parts From Image Collections
Chun-Han Yao, Wei-Chih Hung, Varun Jampani, Ming-Hsuan Yang
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 Part-VAE part prior to 3-mesh model (ablation) | LPD · ShapeNet (13 classes, all) | voxel IoU: 49.0 -> 52.9 | Improved |
| add view adversarial learning (ablation) | LPD · ShapeNet (13 classes, all) | voxel IoU: 51.8 -> 52.9 | Improved |
| add part adversarial learning (ablation) | LPD · ShapeNet (13 classes, all) | voxel IoU: 52.2 -> 52.9 | Improved |
| add color reconstruction (ablation) | LPD · ShapeNet (13 classes, all) | voxel IoU: 52.3 -> 52.9 | Improved |
| use latent Part-VAE parts instead of free-form meshes (ablation) | LPD · ShapeNet (13 classes, all) | voxel IoU: 48.1 -> 52.9 | Improved |
| use latent Part-VAE parts instead of cuboid primitives (ablation) | LPD · ShapeNet (13 classes, all) | voxel IoU: 39.7 -> 52.9 | Improved |
| use latent Part-VAE parts instead of free-form meshes (ablation) | LPD · PartNet chair parts (111 samples) | voxel IoU: 15.5 -> 30.9 | Improved |
| use latent Part-VAE parts instead of cuboids (ablation) | LPD · PartNet chair parts (111 samples) | voxel IoU: 18.8 -> 30.9 | Improved |
| use latent Part-VAE parts instead of free-form meshes (ablation) | LPD · Pascal 3D+ (avg) | voxel IoU: 48.0 -> 49.4 | Improved |
| increase number of parts k from 1 to 6 (ablation) | LPD · ShapeNet Airplane (single-class training) | voxel IoU: 54.5 -> 55.9 | Improved |
| increase number of parts k from 3 to 6 (ablation) | LPD · ShapeNet Chair (single-class training) | voxel IoU: 46.6 -> 46.4 | No clear effect |
| increase number of parts k from 2 to 6 (ablation) | LPD · ShapeNet Car (single-class training) | voxel IoU: 76.1 -> 75.2 | Hurt |
Closest papers by meaning
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.91)
- Learning to Group: A Bottom-Up Framework for 3D Part Discovery in Unseen Categories (ICLR 2020 · similarity 0.90)
- PartNet: A Large-Scale Benchmark for Fine-Grained and Hierarchical Part-Level 3D Object Understanding (CVPR 2019 · similarity 0.90)
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.89)
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.88)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · 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.
