Box2Rip: Instance Segmentation of Amorphous Rip Currents via Box-Supervised Learning
Rip currents are hazardous offshore water flows that significantly threaten swimmers and bathers, pulling them away from the shore at velocities up to <inline-formula> <tex-math notation="LaTeX">$2.4\ ms^{-1}$ </tex-math></inline-formula>. Due to their amorphous str...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11030606/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849417636874551296 |
|---|---|
| author | Juno Choi Muralidharan Rajendran Yong-Cheol Suh |
| author_facet | Juno Choi Muralidharan Rajendran Yong-Cheol Suh |
| author_sort | Juno Choi |
| collection | DOAJ |
| description | Rip currents are hazardous offshore water flows that significantly threaten swimmers and bathers, pulling them away from the shore at velocities up to <inline-formula> <tex-math notation="LaTeX">$2.4\ ms^{-1}$ </tex-math></inline-formula>. Due to their amorphous structure and indistinct boundaries, accurately identifying rip currents remains a challenging task. While deep learning (DL)-based segmentation models have shown promise, they typically rely on laborious and resource-intensive pixel-level annotations. In this paper, we address this crucial problem by introducing Box2Rip, a novel box-supervised instance segmentation (BSIS) framework designed for rip current segmentation using only bounding box annotations. Box2Rip employs a two-phase architecture encompassing an extent-sensitive mask generator, mask refiner, and bbox matcher. Besides, we propose a joint loss system to encourage the predicted masks to adhere to the bounding boxes while capturing rip current boundaries. Our extensive experiments on the Dumitriu dataset with 2,466 images show that Box2Rip with ResNet-50 backbone achieves 76.2% <inline-formula> <tex-math notation="LaTeX">$AP_{50:95}$ </tex-math></inline-formula>, outperforming fully supervised models by 66.3% and exceeding box-supervised methods by 9.4%. On the AI-Hub CCTV rip current dataset containing 30,000 images, Box2Rip attained 38.6% <inline-formula> <tex-math notation="LaTeX">$AP_{50:95}$ </tex-math></inline-formula>, 76.2% <inline-formula> <tex-math notation="LaTeX">$AP_{50}$ </tex-math></inline-formula>, and 60.2% AR with minimal bounding box supervision. The proposed framework demonstrates exceptional capability in delineating amorphous rip currents with significantly reduced annotation effort, enabling efficient and scalable rip current detection systems that enhance coastal safety measures. |
| format | Article |
| id | doaj-art-74b74379e99e4cb69098aa72e50f8b30 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-74b74379e99e4cb69098aa72e50f8b302025-08-20T03:32:42ZengIEEEIEEE Access2169-35362025-01-011310727710729210.1109/ACCESS.2025.357887811030606Box2Rip: Instance Segmentation of Amorphous Rip Currents via Box-Supervised LearningJuno Choi0https://orcid.org/0009-0005-3877-3740Muralidharan Rajendran1https://orcid.org/0009-0006-7577-773XYong-Cheol Suh2https://orcid.org/0000-0003-1871-3044UR Interdisciplinary Program of Hydrography, Pukyong National University, Busan, South KoreaResearch and Development Team, GreenBlue Inc., Busan, South KoreaDepartment of Civil Engineering, Pukyong National University, Busan, South KoreaRip currents are hazardous offshore water flows that significantly threaten swimmers and bathers, pulling them away from the shore at velocities up to <inline-formula> <tex-math notation="LaTeX">$2.4\ ms^{-1}$ </tex-math></inline-formula>. Due to their amorphous structure and indistinct boundaries, accurately identifying rip currents remains a challenging task. While deep learning (DL)-based segmentation models have shown promise, they typically rely on laborious and resource-intensive pixel-level annotations. In this paper, we address this crucial problem by introducing Box2Rip, a novel box-supervised instance segmentation (BSIS) framework designed for rip current segmentation using only bounding box annotations. Box2Rip employs a two-phase architecture encompassing an extent-sensitive mask generator, mask refiner, and bbox matcher. Besides, we propose a joint loss system to encourage the predicted masks to adhere to the bounding boxes while capturing rip current boundaries. Our extensive experiments on the Dumitriu dataset with 2,466 images show that Box2Rip with ResNet-50 backbone achieves 76.2% <inline-formula> <tex-math notation="LaTeX">$AP_{50:95}$ </tex-math></inline-formula>, outperforming fully supervised models by 66.3% and exceeding box-supervised methods by 9.4%. On the AI-Hub CCTV rip current dataset containing 30,000 images, Box2Rip attained 38.6% <inline-formula> <tex-math notation="LaTeX">$AP_{50:95}$ </tex-math></inline-formula>, 76.2% <inline-formula> <tex-math notation="LaTeX">$AP_{50}$ </tex-math></inline-formula>, and 60.2% AR with minimal bounding box supervision. The proposed framework demonstrates exceptional capability in delineating amorphous rip currents with significantly reduced annotation effort, enabling efficient and scalable rip current detection systems that enhance coastal safety measures.https://ieeexplore.ieee.org/document/11030606/Rip current segmentationinstance segmentationbox-supervised learningtwo-phase architectureextent-sensitive mask generation |
| spellingShingle | Juno Choi Muralidharan Rajendran Yong-Cheol Suh Box2Rip: Instance Segmentation of Amorphous Rip Currents via Box-Supervised Learning IEEE Access Rip current segmentation instance segmentation box-supervised learning two-phase architecture extent-sensitive mask generation |
| title | Box2Rip: Instance Segmentation of Amorphous Rip Currents via Box-Supervised Learning |
| title_full | Box2Rip: Instance Segmentation of Amorphous Rip Currents via Box-Supervised Learning |
| title_fullStr | Box2Rip: Instance Segmentation of Amorphous Rip Currents via Box-Supervised Learning |
| title_full_unstemmed | Box2Rip: Instance Segmentation of Amorphous Rip Currents via Box-Supervised Learning |
| title_short | Box2Rip: Instance Segmentation of Amorphous Rip Currents via Box-Supervised Learning |
| title_sort | box2rip instance segmentation of amorphous rip currents via box supervised learning |
| topic | Rip current segmentation instance segmentation box-supervised learning two-phase architecture extent-sensitive mask generation |
| url | https://ieeexplore.ieee.org/document/11030606/ |
| work_keys_str_mv | AT junochoi box2ripinstancesegmentationofamorphousripcurrentsviaboxsupervisedlearning AT muralidharanrajendran box2ripinstancesegmentationofamorphousripcurrentsviaboxsupervisedlearning AT yongcheolsuh box2ripinstancesegmentationofamorphousripcurrentsviaboxsupervisedlearning |