Research map / Point clouds / Point cloud classification
Group Contextual Encoding for 3D Point Clouds
Xu Liu, Chengtao Li, Jian Wang, Jingbo Wang, Boxin Shi, Xiaodong He
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 |
|---|---|---|---|
| Replace SA2 with SA2' using global average pooling (K=0, G=1) (ablation) | PointNet++ VoteNet · ScanNet V2 | mAP@0.25: 51.2 -> 51.7 | Improved |
| Add encoding layer with K=8 code words (G=1) (ablation) | SA2 · SUN RGB-D V1 | mAP@0.25: 51.2 -> 54.6 | Improved |
| Increase code words from K=8 to K=32 (G=1) (ablation) | SA2' · ScanNet V2 | mAP@0.25: 53.0 -> 53.4 | No clear effect |
| Increase code words from K=8 to K=24 (G=1) (ablation) | SA2' · SUN RGB-D V1 | mAP@0.25: 54.6 -> 55.5 | Improved |
| Multiply channel number by 2 without encoding (ablation) | SA2 · SUN RGB-D | mAP@0.25: 51.2 -> 54.0 | Improved |
| Multiply channel number by 3 without encoding (ablation) | SA2 · SUN RGB-D | mAP@0.25: 51.2 -> 54.2 | Improved |
| Replace encoding layer (G=1) with group division (G=12) (ablation) | SA2' C x 3 · SUN RGB-D | mAP@0.25: 55.8 -> 57.1 | Improved |
| Add encoding layer (G=1) on C x 3 baseline (ablation) | SA2 C x 3 · SUN RGB-D | mAP@0.25: 54.2 -> 55.8 | Improved |
| Add channel shuffle to grouping (ablation) | Group contextual encoding C x 3 · SUN RGB-D | mAP@0.25: 57.1 -> 56.3 | Hurt |
| Group contextual encoding vs original SA layer at C x 3 (ablation) | PointNet++ · SUN RGB-D | mAP@0.25: 54.2 -> 57.1 | Improved |
| Add group contextual encoding blocks to VoteNet | VoteNet · ScanNet V2 | mAP@0.5: 33.54 -> 40.11 | Improved |
| Add group contextual encoding blocks to VoteNet | VoteNet · ScanNet V2 | mAP@0.25: 58.65 -> 60.83 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PointWeb — PointWeb: Enhancing Local Neighborhood Features for Point Cloud Processing (CVPR 2019) · better on 1 of 1 shared result
- PointCNN — PointCNN: Convolution On X-Transformed Points (NeurIPS 2018) · better on 1 of 1 shared result
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
- Clustering-based Supervised Training for 3D Perception (TPAMI 2026 · similarity 0.89)
- PointContrast: Unsupervised Pre-training for 3D Point Cloud Understanding (ECCV 2020 · similarity 0.89)
- Explore In-Context Learning for 3D Point Cloud Understanding (NeurIPS 2023 · similarity 0.89)
- Geometrically-driven Aggregation for Zero-shot 3D Point Cloud Understanding (CVPR 2024 · similarity 0.89)
- More Text, Less Point: Towards 3D Data-Efficient Point-Language Understanding (AAAI 2025 · similarity 0.88)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 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.
