Research map / Point clouds / Segmentation and detection
3D Recurrent Neural Networks with Context Fusion for Point Cloud Semantic Segmentation
Xiaoqing Ye, Jiamao Li, Hexiao Huang, Liang Du, Xiaolin Zhang
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 pointwise pyramid pooling (ablation) | PointNet · S3DIS (6-fold CV) | overall accuracy: 78.5 -> 82.8 | Improved |
| add pointwise pyramid pooling (ablation) | PointNet · S3DIS (6-fold CV) | mIoU: 47.6 -> 50.8 | Improved |
| add one-direction RNNs (ablation) | PointNet · S3DIS (6-fold CV) | mIoU: 47.6 -> 49.9 | Improved |
| add two-direction RNNs instead of one-direction (ablation) | PointNet · S3DIS (6-fold CV) | mIoU: 49.9 -> 51.4 | Improved |
| add two-direction RNNs (ablation) | PointNet · S3DIS (6-fold CV) | mAcc: 66.2 -> 70.0 | Improved |
| combine pyramid pooling with two-direction RNNs (ablation) | PointNet · S3DIS (6-fold CV) | mIoU: 50.8 -> 56.3 | Improved |
| combine pyramid pooling with two-direction RNNs (ablation) | PointNet · S3DIS (6-fold CV) | overall accuracy: 78.5 -> 86.9 | Improved |
| increase RNN time step from 1 to 4 (ablation) | full model · S3DIS (6-fold CV) | mIoU: 51.3 -> 56.4 | Improved |
| increase RNN time step from 4 to 10 (ablation) | full model · S3DIS (6-fold CV) | mIoU: 56.4 -> 54.0 | Hurt |
| use the proposed model instead of PointNet++ (XYZ only) | point cloud segmentation · ScanNet | overall accuracy: 0.743 -> 0.765 | Improved |
| use the proposed model instead of G+RCU | point cloud segmentation · vKITTI (XYZ-RGB) | mIoU: 0.362 -> 0.416 | Improved |
| use the proposed model instead of SPG | point cloud segmentation · S3DIS Area 5 | mIoU: 50.7 -> 51.0 | No clear effect |
Closest papers by meaning
- Recurrent Slice Networks for 3D Segmentation of Point Clouds (CVPR 2018 · similarity 0.89)
- 3DCNN-DQN-RNN: A Deep Reinforcement Learning Framework for Semantic Parsing of Large-Scale 3D Point Clouds (ICCV 2017 · similarity 0.89)
- Geometry and Uncertainty-Aware 3D Point Cloud Class-Incremental Semantic Segmentation (CVPR 2023 · similarity 0.89)
- GrowSP: Unsupervised Semantic Segmentation of 3D Point Clouds (CVPR 2023 · similarity 0.88)
- P-SLCR: Unsupervised Point Cloud Semantic Segmentation via Prototypes Structure Learning and Consistent Reasoning (AAAI 2026 · similarity 0.88)
- Weakly Supervised 3D Scene Segmentation with Region-Level Boundary Awareness and Instance Discrimination (ECCV 2022 · similarity 0.88)
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.
