Research map / Label-efficient and robust learning / Domain adaptation and generalization
BiPA: Bilevel Prompt Adaptation for Underwater Instance Segmentation
Long Ma, Haoze Zheng, Yuhang Mao, Jinyuan Liu, Chengpei Xu, Xinwei Xue, Yi Wang, Xiangjian He, Weimin Wang
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 |
|---|---|---|---|
| adapt the dense prompt with bilevel optimization instead of keeping it frozen (ablation) | underwater SAM · UIIS | mAP: 30.9 -> 32.1 | Improved |
| naive joint prompt and model training instead of frozen prompt (ablation) | underwater SAM · USIS10K (multi-class) | mAP: 44.4 -> 43.6 | Hurt |
| bilevel prompt adaptation instead of naive joint training (ablation) | underwater SAM · USIS10K (multi-class) | mAP: 43.6 -> 45.2 | Improved |
| add fine-tuning stage with frozen adapted prompt after adaptation stage (ablation) | BiPA · UIIS | mAP: 29.8 -> 32.1 | Improved |
| use FAI block instead of simple addition fusion of ResNet50 and ViT features (ablation) | BiPA · USIS10K (multi-class) | mAP: 43.5 -> 45.2 | Improved |
| add FAI block under the frozen prompt strategy (ablation) | underwater SAM · UIIS | mAP: 29.5 -> 30.9 | Improved |
| increase optimization interval to B=[1,3] (ablation) | BiPA prompt adaptation · UIIS | mAP: 30.1 -> 32.1 | Improved |
| use 16 prompt groups instead of 8 or 32 (ablation) | BiPA prompt adaptation · UIIS | mAP: 31.5 (Ng=8), 30.8 (Ng=32) -> 32.1 | Improved |
| BiPA versus USIS-SAM on small objects | BiPA · UIIS | APs: 8.1 -> 7.2 | Hurt |
| add ResNet50 injection and bilevel adaptation to USIS-SAM | USIS-SAM · UIIS and USIS10K | parameters and FLOPs: +5.59% params, +0.42% FLOPs, +6.95% average mAP | Mixed |
Closest papers by meaning
- Diving into Underwater: Segment Anything Model Guided Underwater Salient Instance Segmentation and A Large-scale Dataset (ICML 2024 · similarity 0.89)
- NAUTILUS: A Large Multimodal Model for Underwater Scene Understanding (NeurIPS 2025 · similarity 0.84)
- A Unified Image-Dense Annotation Generation Model for Underwater Scenes (CVPR 2025 · similarity 0.83)
- WaterDiffusion: Learning a Prior-involved Unrolling Diffusion for Joint Underwater Saliency Detection and Visual Restoration (AAAI 2025 · similarity 0.83)
- Atlantis: Enabling Underwater Depth Estimation with Stable Diffusion (CVPR 2024 · similarity 0.83)
- Synergistic Multiscale Detail Refinement via Intrinsic Supervision for Underwater Image Enhancement (AAAI 2024 · similarity 0.82)
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.
