arXiv cs.AI· Ping Guo, Chengzhou Li, Guanchen Meng, Qi Jia, Jinyuan Liu, Zhu Liu, Yu Liu, Zhongxuan Luo, Xin Fan·· 4 小时前AI 评分32
CTFS：面向前视声呐图像语义分割的协作教师框架，极少量标注下提升 mIoU
CTFS : Collaborative Teacher Framework for Forward-Looking Sonar Image Semantic Segmentation with Extremely Limited Labels
AI 导读
研究者提出面向前视声呐图像语义分割的协作教师框架 CTFS，采用一个通用教师与多个声呐专属教师的多教师交替指导策略，并设计跨教师可靠性评估机制量化伪标签可靠性以抑制噪声伪标签。在 FLSMD 数据集仅 2% 标注的条件下，该方法 mIoU 较其他 SOTA 方法提升 5.08%，论文已被 CVPR 2026 Finding 接收。
正文 · 原文
Abstract:As one of the most important underwater sensing technologies, forward-looking sonar exhibits unique imaging characteristics. Sonar images are often affected by severe speckle noise, low texture contrast, acoustic shadows, and geometric distortions. These factors make it difficult for traditional teacher-student frameworks to achieve satisfactory performance in sonar semantic segmentation tasks under extremely limited labeled data conditions. To address this issue, we propose a Collaborative Teacher Semantic Segmentation Framework for forward-looking sonar images. This framework introduces a multi-teacher collaborative mechanism composed of one general teacher and multiple sonar-specific teachers. By adopting a multi-teacher alternating guidance strategy, the student model can learn general semantic representations while simultaneously capturing the unique characteristics of sonar images, thereby achieving more comprehensive and robust feature modeling. Considering the challenges of sonar images, which can lead teachers to generate a large number of noisy pseudo-labels, we further design a cross-teacher reliability assessment mechanism. This mechanism dynamically quantifies the reliability of pseudo-labels by evaluating the consistency and stability of predictions across multiple views and multiple teachers, thereby mitigating the negative impact caused by noisy pseudo-labels. Notably, on the FLSMD dataset, when only 2% of the data is labeled, our method achieves a 5.08% improvement in mIoU compared to other state-of-the-art approaches.
| Comments: | Accepted to CVPR 2026 Finding. Code: this https URL |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2603.21071 [cs.CV] |
| (or arXiv:2603.21071v4 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2603.21071 arXiv-issued DOI via DataCite |
Submission history
From: Ping Guo [view email]
[v1]
Sun, 22 Mar 2026 05:57:36 UTC (2,811 KB)
[v2]
Sat, 16 May 2026 13:16:55 UTC (2,811 KB)
[v3]
Mon, 19 Oct 2026 16:53:21 UTC (2,811 KB)
[v4]
Thu, 5 Nov 2026 16:30:56 UTC (2,811 KB)
来源：arXiv cs.AI · arxiv.org