Research map / Point clouds / Point cloud classification
Learning Geometry-Disentangled Representation for Complementary Understanding of 3D Object Point Cloud
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 both sharp and gentle component attention to KNN aggregation (ablation) | GDANet · ModelNet40 | accuracy: 92.2 -> 93.4 | Helped |
| add sharp and gentle components together rather than either alone (ablation) | GDANet with KNN aggregation · ModelNet40 | accuracy: 92.7, 92.4 -> 93.4 | Helped |
| use self-attention on the input point cloud instead of sharp-gentle attention (ablation) | GDANet · ModelNet40 | accuracy: 92.6 -> 92.6 | Hurt |
| replace geometry-based selection with random selection (ablation) | GDM · ModelNet40 | accuracy: 92.6 -> 93.8 | Hurt |
| replace geometry-based selection with FPS (ablation) | GDM · ModelNet40 | accuracy: 92.7 -> 93.8 | Hurt |
| replace geometry-based selection with random selection (ablation) | GDM · ScanObjectNN OBJ_BG | accuracy: 84.3 -> 86.6 | Hurt |
| replace geometry-based selection with FPS (ablation) | GDM · ScanObjectNN OBJ_BG | accuracy: 84.3 -> 86.6 | Hurt |
| calculate adjacency dynamically on learned semantic features instead of precomputing it on coordinates (ablation) | GDM · ModelNet40 | accuracy: 93.0 -> 93.8 | Helped |
| select 512 rather than 1024 points in GDM (ablation) | GDANet · ModelNet40 | accuracy: 92.6 -> 93.2 | Helped |
| select 256 rather than 1024 points in GDM (ablation) | GDANet · ModelNet40 | accuracy: 92.6 -> 93.8 | Helped |
| select 128 rather than 256 points in GDM (ablation) | GDANet · ModelNet40 | accuracy: 93.8 -> 92.9 | Hurt |
| apply voting during testing (ablation) | GDANet · ModelNet40 | accuracy: 93.4 -> 93.8 | Helped |
Closest papers by meaning
- DOCTR: Disentangled Object-Centric Transformer for Point Scene Understanding (AAAI 2024 · similarity 0.86)
- Neural Convolutional Surfaces (CVPR 2022 · similarity 0.86)
- HGNet: Learning Hierarchical Geometry From Points, Edges, and Surfaces (CVPR 2023 · similarity 0.86)
- Convolution in the Cloud: Learning Deformable Kernels in 3D Graph Convolution Networks for Point Cloud Analysis (CVPR 2020 · similarity 0.86)
- DeCoTR: Enhancing Depth Completion with 2D and 3D Attentions (CVPR 2024 · similarity 0.85)
- On Geometry-Enhanced Parameter-Efficient Fine-Tuning for 3D Scene Segmentation (NeurIPS 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.