Research map / Continual learning / Class-incremental
Foundation Model Drives Weakly Incremental Learning for Semantic Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace image-level-only supervision with FMWISS's foundation-model pseudo masks, teacher-student optimization, and memory replay | WILSS segmentation · 15-5 VOC | mIoU (All): 67.3 -> 70.7; 67.2 -> 73.3 | Helped |
| Replace image-level-only supervision with FMWISS | WILSS segmentation · 10-10 VOC | mIoU (All): 60.8 -> 64.6; 65.0 -> 69.1 | Helped |
| Replace WILSON with FMWISS | COCO-to-VOC incremental segmentation · COCO-to-VOC | mIoU (61-80, VOC): 55.7 -> 63.6 | Helped |
| Add union fusion of CLIP masks and seed-guided self-supervised attention masks (ablation) | FMWISS · 10-10 VOC | mIoU (11-20): 53.25 -> 55.45; 56.39 -> 60.54 | Helped |
| Use intersection instead of union for mask fusion (ablation) | FMWISS · 10-10 VOC | mIoU (11-20): 53.25 -> 51.17; 56.39 -> 54.02 | Hurt |
| Add dense contrastive loss with weight 0.1 (ablation) | FMWISS teacher · 10-10 VOC | mIoU (11-20): 55.45 -> 57.22; 60.54 -> 60.92 | Helped |
| Increase dense contrastive loss weight from 0.1 to 1.0 (ablation) | FMWISS teacher · 10-10 VOC | mIoU (All): 63.91 -> 63.87; 68.28 -> 68.17 | Mixed |
| Add memory-based copy-paste augmentation with memory bank size 50 (ablation) | FMWISS · 10-10 VOC | mIoU (1-10): 67.94 -> 68.64; 73.44 -> 73.71 | Helped |
| Replace iBOT ViT-B with iBOT ViT-L (ablation) | FMWISS · 10-10 VOC | mIoU (All): 64.56 -> 64.64; 68.95 -> 69.09 | Mixed |
| Replace iBOT ViT-B with DINO ViT-B (ablation) | FMWISS · 10-10 VOC | mIoU (All): 64.56 -> 64.86; 68.95 -> 68.14 | Mixed |
| Set alpha to 0.5 to combine hard and soft new-class teacher labels (ablation) | FMWISS · 10-10 VOC disjoint | mIoU (All): 62.15 -> 64.21 | Helped |
| Set old-class supervision weight beta to 0.9 (ablation) | FMWISS · 10-10 VOC disjoint | mIoU (All): 59.55 -> 64.56 | Helped |
Closest papers by meaning
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.90)
- Deconfound Semantic Shift and Incompleteness in Incremental Few-shot Semantic Segmentation (AAAI 2025 · similarity 0.89)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.88)
- Separate and Conquer: Decoupling Co-occurrence via Decomposition and Representation for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.88)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.88)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic 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.