Research map / Detection and segmentation / CNN features
SNE-RoadSeg: Incorporating Surface Normal Information into Semantic Segmentation for Accurate Freespace Detection
Rui Fan, Hengli Wang, Peide Cai, Ming Liu
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 surface normal information from SNE to depth input (ablation) | RoadSeg-152 depth only · R2D road dataset | IoU (%): 92.4 -> 94.5 | Improved |
| add SNE to four-channel RGB-D input (ablation) | RoadSeg-152 single encoder · R2D road dataset | IoU (%): 89.0 -> 93.0 | Improved |
| use two-encoder fusion instead of four-channel input (ablation) | RoadSeg-152 with SNE · R2D road dataset | IoU (%): 93.0 -> 96.7 | Improved |
| add depth encoder to RGB-only model (ablation) | RoadSeg-152 · R2D road dataset | IoU (%): 86.6 -> 96.7 | Improved |
| use ResNet-152 instead of ResNet-18 backbone (ablation) | RoadSeg · R2D road dataset | IoU (%): 85.9 -> 96.7 | Improved |
| use ResNet-152 instead of ResNet-101 backbone (ablation) | RoadSeg · R2D road dataset | F-score (%): 97.6 -> 98.3 | Improved |
| add dense skip connections in decoder instead of none (ablation) | RoadSeg-152 · R2D road dataset | IoU (%): 95.2 -> 96.7 | Improved |
| use dense skip connections instead of U-Net-style sparse ones (ablation) | RoadSeg-152 · R2D road dataset | IoU (%): 95.9 -> 96.7 | Improved |
| embed SNE into single-modal CNNs | ten state-of-the-art CNNs · SYNTHIA and R2D | IoU (%): approximately 2-12% | Improved |
| embed SNE into data-fusion CNNs | data-fusion CNNs · SYNTHIA and R2D | IoU (%): about 1-7% | Improved |
| use RoadSeg instead of other CNNs, all with SNE | RoadSeg with SNE · SYNTHIA and R2D | IoU (%): approximately 1.4-14.7% | Improved |
| compare SNE-RoadSeg with RBANet on KITTI | SNE-RoadSeg · KITTI road benchmark | MaxF (%): 96.30 -> 96.75 | Improved |
Closest papers by meaning
- In Defense of Pre-Trained ImageNet Architectures for Real-Time Semantic Segmentation of Road-Driving Images (CVPR 2019 · similarity 0.85)
- Efficient Outdoor 3D Point Cloud Semantic Segmentation for Critical Road Objects and Distributed Contexts (ECCV 2020 · similarity 0.85)
- GSENet:Global Semantic Enhancement Network for Lane Detection (AAAI 2024 · similarity 0.85)
- ViiNeuS: Volumetric Initialization for Implicit Neural Surface Reconstruction of Urban Scenes with Limited Image Overlap (CVPR 2025 · similarity 0.84)
- DriveGEN: Generalized and Robust 3D Detection in Driving via Controllable Text-to-Image Diffusion Generation (CVPR 2025 · similarity 0.83)
- Weakly Supervised Semantic Segmentation for Driving Scenes (AAAI 2024 · similarity 0.83)
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.
