Research map / Label-efficient and robust learning / Domain adaptation and generalization
Adapting In-Domain Few-Shot Segmentation to New Domains without Source Domain Retraining
Qi Fan, Kaiqi Liu, Nian Liu, Hisham Cholakkal, Rao Muhammad Anwer, Wenbin Li, Yang Gao
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 informative structure adaptation (ISA) on top of a well-trained in-domain FSS model | SSP ResNet-50 · CD-FSS 4 datasets (5-shot avg) | mIoU: 63.0 -> 70.3 | Improved |
| add ISA to PANet, FPTrans, PerSAM and DR-Adapter | various FSS and CD-FSS models · CD-FSS 4 datasets (5-shot avg) | mIoU: +6.4 to +7.3 (DR-Adapter +5.5) | Improved |
| add model structure adaptation baseline (ablation) | SSP ResNet-50 · CD-FSS 4 datasets (5-shot avg) | mIoU: 63.0 -> 64.6 | Improved |
| add informative structure identification (ISI) (ablation) | MSA SSP · CD-FSS 4 datasets (5-shot avg) | mIoU: 64.6 -> 66.5 | Improved |
| add progressive structure adaptation (PSA) (ablation) | MSA SSP · CD-FSS 4 datasets (5-shot avg) | mIoU: 64.6 -> 67.6 | Improved |
| add ISI on top of MSA + PSA (ablation) | ISA · Deepglobe | mIoU: 53.2 -> 52.7 | Mixed |
| select trainable layers with ISI instead of training only the last conv layer (ablation) | ISA · CD-FSS 4 datasets (5-shot avg) | mIoU: 67.6 -> 70.3 | Improved |
| use largest structure Fisher score per layer instead of the mean (ablation) | ISI · CD-FSS 4 datasets (5-shot avg) | mIoU: 67.8 -> 70.3 | Improved |
| increase ISI trainable layers from 1 to 3 (ablation) | ISI · CD-FSS 4 datasets (5-shot avg) | mIoU: 70.3 -> 71.0 | No clear effect |
| use progressive training 1->2->3->4 instead of directly training with 4-shot samples (ablation) | PSA · CD-FSS 4 datasets (5-shot avg) | mIoU: 67.8 -> 70.3 | Improved |
| increase support shots from 5 to 30 (ablation) | ISA vs SSP baseline · CD-FSS 4 datasets | mIoU: 70.3 -> 74.6 (baseline 63.0 -> 65.2) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- APSeg — APSeg: Auto-Prompt Network for Cross-Domain Few-Shot Semantic Segmentation (CVPR 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Adapt Before Comparison: A New Perspective on Cross-Domain Few-Shot Segmentation (CVPR 2024 · similarity 0.94)
- Domain-Rectifying Adapter for Cross-Domain Few-Shot Segmentation (CVPR 2024 · similarity 0.92)
- Adapter Naturally Serves as Decoupler for Cross-Domain Few-Shot Semantic Segmentation (ICML 2025 · similarity 0.92)
- The Devil is in Low-Level Features for Cross-Domain Few-Shot Segmentation (CVPR 2025 · similarity 0.90)
- Cross-Domain Few-Shot Segmentation via Iterative Support-Query Correspondence Mining (CVPR 2024 · similarity 0.90)
- Dual-Agent Optimization framework for Cross-Domain Few-Shot Segmentation (CVPR 2025 · similarity 0.90)
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.
