Research map / Point clouds / Segmentation and detection
3D Semantic Segmentation With Submanifold Sparse Convolutional Networks
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 |
|---|---|---|---|
| Use submanifold sparse convolutions instead of the compared baseline approaches | ShapeNet random-pose segmentation models · ShapeNet test set | Average IoU at 10^8 FLOPs: 6-8% | Helped |
| Add downsampling and upsampling operations (ablation) | SSCN architecture, compared with single-resolution C3 · ShapeNet test set | Average IoU | Helped |
| Increase voxelization scale (ablation) | SSCN C3, FCN, and U-Net · ShapeNet test set | Average IoU | Mixed |
| Use four-view rather than one-view testing (ablation) | SSCN-FCN A · NYU Depth v2 test set | Pixel accuracy: 64.1% -> 66.9% | Helped |
| Use four-view rather than one-view testing (ablation) | SSCN-FCN B · NYU Depth v2 test set | Pixel accuracy: 66.4% -> 68.5% | Helped |
| Remove depth information by setting depth values to zero (ablation) | SSCN-FCN A · NYU Depth v2 test set | Pixel accuracy: 64.1% to 50.8% | Hurt |
| Randomly rotate the objects rather than retain axis alignment (ablation) | Shape-context classifier · ShapeNet | IoU accuracy: 63.5% -> 47.8% | Hurt |
Closest papers by meaning
- OA-CNNs: Omni-Adaptive Sparse CNNs for 3D Semantic Segmentation (CVPR 2024 · similarity 0.90)
- SBNet: Sparse Blocks Network for Fast Inference (CVPR 2018 · similarity 0.89)
- Spatial Pruned Sparse Convolution for Efficient 3D Object Detection (NeurIPS 2022 · similarity 0.89)
- 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks (CVPR 2019 · similarity 0.89)
- SpSequenceNet: Semantic Segmentation Network on 4D Point Clouds (CVPR 2020 · similarity 0.89)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 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.