Research map / Detection and segmentation / Weakly supervised detection and segmentation
CoralSRT: Revisiting Coral Reef Semantic Segmentation by Feature Rectification via Self-supervised Guidance
Ziqiang Zheng, Yuk-Kwan Wong, Binh-Son Hua, Jianbo Shi, Sai-Kit Yeung
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 CoralSRT feature rectification | DINOv2 · Sparse-to-dense, 100 points | mIoU: 60.23 -> 63.40 | Improved |
| add CoralSRT feature rectification | SAM · Sparse-to-dense, 100 points | mIoU: 58.27 -> 63.14 | Improved |
| add CoralSRT feature rectification | SAM 2 · Sparse-to-dense, 100 points | mIoU: 58.39 -> 61.24 | Improved |
| add CoralSRT feature rectification | CoralSCOP · Sparse-to-dense, 100 points | mIoU: 58.97 -> 65.46 | Improved |
| label propagation in feature space instead of point prompts | SAM 2 · Sparse-to-dense, 100 points | mIoU: 28.13 -> 58.39 | Improved |
| replace DVT denoising with CoralSRT | DINOv2 · Sparse-to-dense, 5 points | mIoU: 30.02 -> 30.68 | Improved |
| add CoralSRT on top of FeatUp upsampling | FeatUp (DINOv2) · Sparse-to-dense, 100 points | mIoU: 64.32 -> 65.78 | Improved |
| add CoralSRT (ablation) | DINO CoralWorld-0.1M features · Sparse-to-dense, 5 points | mIoU: 25.86 -> 29.47 | Improved |
| scale pretraining data from 0.1M to 2.64M images (ablation) | DINO · Sparse-to-dense, 50 points | mIoU: 50.49 -> 54.43 | Improved |
| add training of Rec instead of test-time rectification with SAM 2 masks (ablation) | CoralSRT (DINOv2) · Sparse-to-dense, 50 points | mIoU: 53.56 -> 57.46 | Improved |
| use median instead of mean centrality (ablation) | CoralSRT (DINOv2) · Sparse-to-dense, 50 points | mIoU: 57.38 -> 57.46 | No clear effect |
| optimize Rec with human-annotated masks instead of SAM 2 masks (ablation) | CoralSRT (DINOv2) · Sparse-to-dense, 50 points | mIoU: 57.46 -> 57.65 | No clear effect |
Closest papers by meaning
- CoralSCOP: Segment any COral Image on this Planet (CVPR 2024 · similarity 0.93)
- Enhancing Vision-Language Models with Morphological and Taxonomic Knowledge: Towards Coral Recognition for Ocean Health (AAAI 2025 · similarity 0.91)
- ReefNet: A Large-Scale Dataset and Benchmark for Fine-Grained Coral Reef Recognition (NeurIPS 2026 · similarity 0.90)
- CoralVQA: A Large-Scale Visual Question Answering Dataset for Coral Reef Image Understanding (NeurIPS 2025 · similarity 0.88)
- SeafloorAI: A Large-scale Vision-Language Dataset for Seafloor Geological Survey (NeurIPS 2024 · similarity 0.86)
- LAqua: Laplacian Pyramids for Aquatic Segmentation (Student Abstract) (AAAI 2025 · similarity 0.84)
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.
