Research map / Point clouds / Segmentation and detection
xMHashSeg: Cross-modal Hash Learning for Training-free Unsupervised LiDAR Semantic Segmentation
Jialong Zhang, Yachao Zhang, Yao Wu, Jiangming Shi, Fangyong Wang, Yanyun Qu
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 |
|---|---|---|---|
| use Point-SANN instead of Point-NN (ablation) | xMHashSeg · nuScenes Night subset | mIoU: 32.1 -> 41.7 | Improved |
| add depth-map view as bridge between 2D and 3D (ablation) | xMHashSeg · nuScenes Night subset | mIoU: 36.9 -> 41.7 | Improved |
| learn unified hash code instead of k-means on concatenated features (ablation) | xMHashSeg · nuScenes Night subset | mIoU: 33.9 -> 41.7 | Improved |
| set balance factor lambda to 0.25 instead of 0 (ablation) | xMHashSeg · nuScenes Night subset | mIoU: 37.1 -> 41.7 | Improved |
| set balance factor lambda to 0.25 instead of 1 (ablation) | xMHashSeg · nuScenes Night subset | mIoU: 37.6 -> 41.7 | Improved |
| training-free method vs best zero-shot method | xMHashSeg · nuScenes (ZSL unseen 6 classes) | mIoU: 22.4 -> 24.4 | Improved |
| training-free method vs best zero-shot method | xMHashSeg · SemanticKITTI (ZSL unseen 6 classes) | mIoU: 23.9 -> 24.2 | Improved |
| training-free method vs UDA methods | xMHashSeg · nuScenes (6 class), UDA/TTA comparison | mIoU: *64.0 -> 40.9 | Hurt |
| training-free inference speed vs Latte | xMHashSeg · SemanticKITTI | seconds per scene: 2.7 s -> 1.5 s per scene | Improved |
Closest papers by meaning
- Unsupervised Semantic Segmentation of Urban Scenes via Cross-Modal Distillation (IJCV 2025 · similarity 0.89)
- Uni-to-Multi Modal Knowledge Distillation for Bidirectional LiDAR-Camera Semantic Segmentation (TPAMI 2024 · similarity 0.88)
- Drive&Segment: Unsupervised Semantic Segmentation of Urban Scenes via Cross-Modal Distillation (ECCV 2022 · similarity 0.88)
- Less Is More: Reducing Task and Model Complexity for 3D Point Cloud Semantic Segmentation (CVPR 2023 · similarity 0.88)
- (AF)2-S3Net: Attentive Feature Fusion With Adaptive Feature Selection for Sparse Semantic Segmentation Network (CVPR 2021 · similarity 0.88)
- UniDSeg: Unified Cross-Domain 3D Semantic Segmentation via Visual Foundation Models Prior (NeurIPS 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.
