Research map / Point clouds / Segmentation and detection
3D Graph Neural Networks for RGBD Semantic Segmentation
Xiaojuan Qi, Renjie Liao, Jiaya Jia, Sanja Fidler, Raquel Urtasun
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 a graph neural network on top of the unary CNN (ablation) | 3DGNN · NYUD2 test (40-class) | mean IoU (%): 37.1 -> 39.9 | Improved |
| build the graph in 3D space instead of 2D pixel positions (ablation) | 3DGNN · NYUD2 test (40-class) | mean IoU (%): 38.7 -> 39.9 | Improved |
| build the graph in 3D space instead of 2D pixel positions (ablation) | 3DGNN · SUN-RGBD test | mean IoU (%): 38.9 -> 40.2 | Improved |
| use a 2D-position graph neural network instead of unary CNN alone (ablation) | 2DGNN · NYUD2 test (40-class) | mean IoU (%): 37.1 -> 38.7 | Improved |
| add 4 propagation steps over no propagation (ablation) | 3DGNN · NYUD2 validation | mean IoU (%): 37.9 -> 39.4 | Improved |
| increase propagation steps from 4 to 6 (ablation) | 3DGNN · NYUD2 validation | mean IoU (%): 39.4 -> 39.0 | Hurt |
| use LSTM update instead of vanilla RNN update (ablation) | 3DGNN · NYUD2 validation | mean IoU (%) | No clear effect |
| add multi-scale testing | 3DGNN · NYUD2 test (40-class) | mean IoU (%): 39.9 -> 41.7 | Improved |
| combine HHA features with the 3DGNN unary model | 3DGNN · NYUD2 test (40-class, single-scale) | mean IoU (%): 39.9 -> 42.0 | Improved |
| replace VGG-16 with ResNet-101 as unary model | 3DGNN · SUN-RGBD test (single-scale) | mean IoU (%): 40.2 -> 44.1 | Improved |
| add 3DGNN on top of ResNet-101 unary model | ResNet-101 · SUN-RGBD test (single-scale) | mean IoU (%): 42.7 -> 44.1 | Improved |
Closest papers by meaning
- Depth-aware CNN for RGB-D Segmentation (ECCV 2018 · similarity 0.89)
- 3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans (CVPR 2019 · similarity 0.89)
- 3DMV: Joint 3D-Multi-View Prediction for 3D Semantic Scene Segmentation (ECCV 2018 · similarity 0.88)
- DFormer++: Improving RGBD Representation Learning for Semantic Segmentation (TPAMI 2026 · similarity 0.88)
- Incremental Class Discovery for Semantic Segmentation With RGBD Sensing (ICCV 2019 · similarity 0.88)
- DFormer: Rethinking RGBD Representation Learning for Semantic Segmentation (ICLR 2024 · 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.
