Research map / 3D vision / 3D generation
Beyond Viewpoint: Robust 3D Object Recognition under Arbitrary Views through Joint Multi-Part Representation
Linlong Fan, Ye Huang, Yanqi Ge, Wen Li, Lixin Duan
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-aware loss L_awe (ablation) | PANet · ScanObjectNN rotated | accuracy: +2.40% / +1.76% | Improved |
| add cross-view association (CVA) module (ablation) | PANet · ScanObjectNN rotated | accuracy: +0.69% ins., +1.23% class | Improved |
| increase number of views from 1 to 20 (ablation) | PANet ResNet50 · ScanObjectNN rotated | accuracy: 67.07% -> 86.62% | Improved |
| use stronger backbone ResNet50 instead of ResNet18 (ablation) | PANet · ScanObjectNN rotated, 20 views | accuracy: 82.85% -> 86.62% | Improved |
| furthest point sampling instead of random view sampling (ablation) | PANet · ModelNet40 / ScanObjectNN | accuracy: 97.53% vs 97.41% (ModelNet aligned ins.) | No clear effect |
| increase attention maps M from 16 to 64 (ablation) | PANet · ScanObjectNN rotated | instance accuracy: 86.28% -> 87.31% | Improved |
| increase attention maps from 64 to 128 (ablation) | PANet · ScanObjectNN rotated | instance accuracy: 87.31% -> 87.14% | No clear effect |
| increase global part tokens from 16 to 32 (ablation) | PANet · ScanObjectNN rotated | instance accuracy: 87.31% -> 86.31% | Hurt |
| use multiple parts (L=16) instead of a single part (ablation) | PANet · ScanObjectNN rotated | instance accuracy: 86.28% -> 87.31% | Improved |
| switch from fixed to arbitrary views | PANet vs CVR · ScanObjectNN rotated | instance accuracy: PANet 87.82% -> 86.45%; CVR 85.59% -> 71.36% | Mixed |
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.
- View-GCN++ — View-GCN: View-Based Graph Convolutional Network for 3D Shape Analysis (CVPR 2020) · 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
- Learning Relationships for Multi-View 3D Object Recognition (ICCV 2019 · similarity 0.88)
- RotationNet: Joint Object Categorization and Pose Estimation Using Multiviews From Unsupervised Viewpoints (CVPR 2018 · similarity 0.88)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.88)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.87)
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.85)
- Duoduo CLIP: Efficient 3D Understanding with Multi-View Images (ICLR 2025 · similarity 0.85)
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.
