Research map / Point clouds / Segmentation and detection
MSeg3D: Multi-Modal 3D Semantic Segmentation for Autonomous Driving
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 |
|---|---|---|---|
| Add asymmetric multi-modal data augmentation instead of LiDAR-only augmentation (ablation) | GF-Phase model · nuScenes val | mIoU: 71.35 -> 72.39 | Helped |
| Add asymmetric multi-modal data augmentation instead of LiDAR-only augmentation (ablation) | GF-Phase model · Waymo val | mIoU: 60.34 -> 63.97 | Helped |
| Add GF-Phase fusion without feature completion or SF-Phase (ablation) | LiDAR-only model · Waymo val | mIoU: 67.48 -> 63.97 | Hurt |
| Add pixel-to-point cross-modal supervision (ablation) | GF-Phase model · nuScenes val | mIoU: 72.39 -> 76.44 | Helped |
| Add pixel-to-point cross-modal supervision (ablation) | GF-Phase model · Waymo val | mIoU: 63.97 -> 67.13 | Helped |
| Add pseudo-camera feature completion for points outside the camera field of view (ablation) | GF-Phase model with pixel-to-point supervision · nuScenes val | mIoU: 76.44 -> 78.28 | Helped |
| Add pseudo-camera feature completion for points outside the camera field of view (ablation) | GF-Phase model with pixel-to-point supervision · Waymo val | mIoU: 67.13 -> 67.89 | Helped |
| Add SF-Phase semantic feature fusion (ablation) | GF-Phase model with completed camera features · nuScenes val | mIoU: 78.28 -> 80.00 | Helped |
| Add SF-Phase semantic feature fusion (ablation) | GF-Phase model with completed camera features · Waymo val | mIoU: 67.89 -> 69.63 | Helped |
| Use an image backbone rather than no image backbone (ablation) | MSeg3D · nuScenes val | mIoU: 72.00 -> 80.00 | Helped |
| Increase available cameras from 6 to 0 (ablation) | MSeg3D · nuScenes | mIoU: 80.00 -> 74.47 | Hurt |
| Use 25 LiDAR frames instead of 1, with no camera input (ablation) | LiDAR-only baseline · nuScenes | mIoU: 72.00 -> 75.77 | Helped |
Closest papers by meaning
- MSMDFusion: Fusing LiDAR and Camera at Multiple Scales With Multi-Depth Seeds for 3D Object Detection (CVPR 2023 · similarity 0.90)
- MFINet: Multi-view Fusion and 2D–3D Interaction Enhancement for Real-Time LiDAR Semantic Segmentation (AAAI 2026 · similarity 0.90)
- SGFormer: Semantic-Geometry Fusion Transformer for Multi-modal 3D Panoptic Segmentation (AAAI 2025 · similarity 0.88)
- Multi-Space Alignments Towards Universal LiDAR Segmentation (CVPR 2024 · similarity 0.88)
- SDGOCC: Semantic and Depth-Guided Bird's-Eye View Transformation for 3D Multimodal Occupancy Prediction (CVPR 2025 · similarity 0.87)
- VLM2Scene: Self-Supervised Image-Text-LiDAR Learning with Foundation Models for Autonomous Driving Scene Understanding (AAAI 2024 · similarity 0.86)
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.