Research map / Label-efficient and robust learning / Semi-supervised learning
RePL: Pseudo-label Refinement for Semi-supervised LiDAR Semantic Segmentation
Donghyeon Kwon, Taegyu Park, Suha Kwak
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 refiner supervised loss L_rsup (ablation) | RePL refiner · nuScenes-lidarseg val (1% labeled) | mIoU: 50.9 -> 57.2 | Improved |
| add negative-learning loss L_runl for refiner (ablation) | RePL refiner · nuScenes-lidarseg val (1% labeled) | mIoU: 57.2 -> 58.7 | Improved |
| add mixed-scene refiner loss L_rmix (ablation) | RePL refiner · nuScenes-lidarseg val (1% labeled) | mIoU: 58.7 -> 60.0 | Improved |
| add symmetric cross-entropy for unlabeled loss (ablation) | RePL student · nuScenes-lidarseg val (1% labeled) | mIoU: 58.0 -> 60.0 | Improved |
| add random masking when training refiner (ablation) | RePL refiner · nuScenes-lidarseg val (1% labeled) | mIoU: 57.7 -> 60.0 | Improved |
| replace random error mask with teacher-student agreement heuristic (ablation) | RePL refinement · nuScenes-lidarseg val (1% labeled) | mIoU: 58.7 -> 60.0 | Improved |
| replace heuristic error mask with oracle mask (ablation) | RePL refinement · nuScenes-lidarseg val (1% labeled) | mIoU: 60.0 -> 67.3 | Improved |
| reconstruct unreliable voxels instead of discarding them (ablation) | RePL · nuScenes-lidarseg val (1% labeled) | mIoU: 53.5 -> 60.0 | Improved |
| set confidence percentile kappa to 0.4 rather than 0.2/0.6 (ablation) | RePL · nuScenes-lidarseg val (1% labeled) | mIoU: 55.1 / 58.4 -> 60.0 | Mixed |
| apply RePL to SphereFormer backbone (ablation) | SphereFormer · nuScenes-lidarseg val (1% labeled) | mIoU: 50.7 -> 56.5 | Improved |
| apply refinement to pseudo-labels | RePL · nuScenes-lidarseg 1% per-class pseudo-label IoU | IoU: 14 of 16 classes improved; driv. surf. -1.47, sidewalk -2.40 | Mixed |
| compare RePL with IT2 on average mIoU | RePL · nuScenes-lidarseg avg | mIoU: 69.3 -> 71.3 | Improved |
Closest papers by meaning
- Collaborative Learning for Semi-Supervised LiDAR Semantic Segmentation (ICML 2026 · similarity 0.91)
- LaserMix for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2023 · similarity 0.91)
- LESS: Label-Efficient Semantic Segmentation for LiDAR Point Clouds (ECCV 2022 · similarity 0.90)
- ItTakesTwo: Leveraging Peer Representations for Semi-supervised LiDAR Semantic Segmentation (ECCV 2024 · similarity 0.89)
- Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2025 · similarity 0.89)
- Beyond the Label Itself: Latent Labels Enhance Semi-supervised Point Cloud Panoptic Segmentation (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.
