Research map / Point clouds / Segmentation and detection
Label-Guided Knowledge Distillation for Continual Semantic Segmentation on 2D Images and 3D Point Clouds
Ze Yang, Ruibo Li, Evan Ling, Chi Zhang, Yiming Wang, Dezhao Huang, Keng Teck Ma, Minhoe Hur, Guosheng Lin
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 |
|---|---|---|---|
| transplant background probability to ground-truth novel class (vs expand only) (ablation) | LGKD+PLOP · VOC 15-5 | novel mIoU: 30.9 -> 56.1 | Improved |
| keep old class probabilities when transplanting instead of one-hot hard labels (ablation) | LGKD+PLOP · VOC 15-5 | novel mIoU: 51.9 -> 56.1 | Improved |
| add novel set distillation (lambda_n=1) to background set distillation (ablation) | LGKD+PLOP · VOC 15-5 | novel mIoU: 52.0 -> 56.1 | Improved |
| increase background distillation weight lambda_b to 10 (ablation) | LGKD+PLOP · VOC 15-5 | novel mIoU: 53.6 -> 49.8 | Hurt |
| add standard KD to PLOP (ablation) | PLOP · VOC 15-1 / ADE20K | novel mIoU: 21.0 -> 13.4 (VOC 15-1) | No clear effect |
| add unbiased KD to PLOP (ablation) | PLOP · ADE20K 100-50 | novel mIoU: 14.6 -> 14.5 | No clear effect |
| add LGKD to PLOP | PLOP · ADE20K 100-50 | novel mIoU: 14.6 -> 25.7 | Improved |
| add LGKD to MiB | MiB · VOC 19-1 | novel mIoU: 23.2 -> 40.4 | Improved |
| add LGKD to MiB on 3D point clouds | MiB · ScanNet 14-1 | novel mIoU: 29.7 -> 38.4 | Improved |
| add LGKD to MiB at 15-1 | MiB · VOC 15-1 | background IoU: 38.0 -> 39.3 background; all 34.5 -> 35.9 | Mixed |
Closest papers by meaning
- Class Similarity Weighted Knowledge Distillation for Continual Semantic Segmentation (CVPR 2022 · similarity 0.92)
- Make a Strong Teacher with Label Assistance: A Novel Knowledge Distillation Approach for Semantic Segmentation (ECCV 2024 · similarity 0.91)
- Geometry and Uncertainty-Aware 3D Point Cloud Class-Incremental Semantic Segmentation (CVPR 2023 · similarity 0.89)
- Preparing the Future for Continual Semantic Segmentation (ICCV 2023 · similarity 0.89)
- Distilling Knowledge from Heterogeneous Architectures for Semantic Segmentation (AAAI 2025 · similarity 0.88)
- TIKP: Text-to-Image Knowledge Preservation for Continual Semantic Segmentation (AAAI 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.
