Research map / Point clouds / Segmentation and detection
EPMF: Efficient Perception-Aware Multi-Sensor Fusion for 3D Semantic Segmentation
Mingkui Tan, Zhuangwei Zhuang, Sitao Chen, Rong Li, Kui Jia, Qicheng Wang, Yuanqing Li
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 perspective projection (vs spherical) with LiDAR-only input (ablation) | SalsaNext baseline · SemanticKITTI-FV val | mIoU: 57.2 -> 57.6 | Improved |
| add perspective projection (vs spherical) in multimodal fusion (ablation) | PMF · SemanticKITTI-FV val | mIoU: 55.8 -> 61.7 | Improved |
| add ASPP to LiDAR stream (ablation) | PMF · SemanticKITTI-FV val | mIoU: 57.6 -> 59.7 | Improved |
| add residual-based fusion modules (ablation) | PMF · SemanticKITTI-FV val | mIoU: 59.7 -> 61.7 | Improved |
| add perception-aware loss (ablation) | PMF · SemanticKITTI-FV val | mIoU: 61.7 -> 63.9 | Improved |
| drop the decoder of the camera stream (ablation) | PMF · SemanticKITTI-FV val | mIoU: 63.9 -> 63.9 | No clear effect |
| add cross-modal alignment and crop (ablation) | PMF · SemanticKITTI-FV val | FLOPs: 854.7 G -> 739.9 G | Improved |
| add improved contextual module (ablation) | EPMF · SemanticKITTI-FV val | mIoU: 63.9 -> 64.8 | Improved |
| fuse high-level LiDAR features into camera stream (ablation) | EPMF · SemanticKITTI-FV val | mIoU: 64.8 -> 65.9 | Improved |
| mask 10% of image from bottom (no fine-tuning) (ablation) | EPMF · SemanticKITTI-FV val | mIoU: 65.9 -> 65.5 | Hurt |
| mask 50% of image (no fine-tuning) (ablation) | EPMF · SemanticKITTI-FV val | mIoU: 65.9 -> 54.8 | Hurt |
| fine-tune with image masks at 50% masked ratio (ablation) | EPMF · SemanticKITTI-FV val | mIoU: 54.8 -> 60.6 | Improved |
Later papers that compared with this method
1 later accepted paper reports EPMF as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Image-to-Point Cloud Feature Back-Projection for Multimodal Training of 3D Semantic Segmentation (CVPR 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- 2DPASS — 2DPASS: 2D Priors Assisted Semantic Segmentation on LiDAR Point Clouds (ECCV 2022) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Perception-Aware Multi-Sensor Fusion for 3D LiDAR Semantic Segmentation (ICCV 2021 · similarity 0.94)
- Exploiting the Benefits of Temporal Information in the Realm of LiDAR Panoptic Segmentation (TPAMI 2025 · similarity 0.89)
- Image-to-Point Cloud Feature Back-Projection for Multimodal Training of 3D Semantic Segmentation (CVPR 2026 · similarity 0.89)
- MFINet: Multi-view Fusion and 2D–3D Interaction Enhancement for Real-Time LiDAR Semantic Segmentation (AAAI 2026 · similarity 0.88)
- PC-BEV: An Efficient Polar-Cartesian BEV Fusion Framework for LiDAR Semantic Segmentation (AAAI 2025 · similarity 0.88)
- Multi-Space Alignments Towards Universal LiDAR Segmentation (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.
