Research map / Point clouds / Segmentation and detection
Unsupervised Semantic Segmentation of Urban Scenes via Cross-Modal Distillation
Antonín Vobecký, David Hurych, Oriane Siméoni, Spyros Gidaris, Andrei Bursuc, Patrick Pérez, Josef Šivic
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 LiDAR-based cross-modal segment extraction instead of FH image segments (ablation) | Drive&Segment (ResNet18+FPN) · Cityscapes | mIoU: 15.5 -> 17.4 | Improved |
| use LiDAR-based cross-modal segment extraction instead of FH image segments (ablation) | Drive&Segment (Segmenter) · Cityscapes | mIoU: 15.8 -> 20.4 | Improved |
| use LiDAR-based segment extraction instead of FH (ablation) | Drive&Segment (Segmenter) · Cityscapes | pixel accuracy: 51.8 -> 65.4 | Improved |
| add teacher-student distillation (ablation) | Drive&Segment (ResNet18+FPN) · Cityscapes | mIoU: 17.4 -> 18.8 | Improved |
| add teacher-student distillation (ablation) | Drive&Segment (Segmenter) · Cityscapes | mIoU: 20.4 -> 20.8 | Improved |
| add teacher-student distillation (ablation) | PiCIE (ResNet18+FPN) · Cityscapes | pixel accuracy: 48.6 -> 64.1 | Improved |
| add LiDAR cross-modal spatial constraints to distillation (ablation) | Drive&Segment (ResNet18+FPN) · Cityscapes | mIoU: 18.8 -> 19.5 | Improved |
| add LiDAR cross-modal spatial constraints to distillation (ablation) | Drive&Segment (Segmenter) · Cityscapes | mIoU: 20.8 -> 21.8 | Improved |
| add LiDAR cross-modal spatial constraints to distillation (ablation) | PiCIE student · Cityscapes | mIoU: 14.8 -> 15.1 | Improved |
| use DeiT instead of DINO features for k-means labeling (ablation) | Drive&Segment (Segmenter) · Cityscapes | mIoU: 20.2 -> 21.7 | Improved |
| downsample LiDAR from 64 to 32 beams (ablation) | Drive&Segment teacher · Cityscapes | mIoU: 20.4 -> 20.3 | No clear effect |
| train on nuScenes instead of Waymo Open | Drive&Segment (Segmenter) · Cityscapes | mIoU: 21.8 -> 19.8 | Hurt |
Closest papers by meaning
- Drive&Segment: Unsupervised Semantic Segmentation of Urban Scenes via Cross-Modal Distillation (ECCV 2022 · similarity 0.99)
- Uni-to-Multi Modal Knowledge Distillation for Bidirectional LiDAR-Camera Semantic Segmentation (TPAMI 2024 · similarity 0.91)
- A Curriculum Domain Adaptation Approach to the Semantic Segmentation of Urban Scenes (TPAMI 2019 · similarity 0.90)
- SemanticKITTI: A Dataset for Semantic Scene Understanding of LiDAR Sequences (ICCV 2019 · similarity 0.89)
- Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2025 · similarity 0.88)
- SUM Parts: Benchmarking Part-Level Semantic Segmentation of Urban Meshes (CVPR 2025 · similarity 0.88)
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.
