This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
Advancing Marine Research: UWSAM Framework and UIIS10K Dataset for Precise Underwater Instance Segmentation
Created by
Haebom
Author
Hua Li, Shijie Lian, Zhiyuan Li, Runmin Cong, Chongyi Li, Laurence T. Yang, Weidong Zhang, Sam Kwong
Outline
To address the challenges of underwater object segmentation, this paper presents the large-scale underwater object instance segmentation dataset UIIS10K (10,048 images, 10 categories) and proposes an efficient underwater object instance segmentation model, UWSAM. UWSAM performs effective visual representation learning using the Mask GAT-based Underwater Knowledge Distillation (MG-UKD) technique, which efficiently distills the knowledge from SAM's ViT-Huge image encoder into a ViT-Small encoder. Furthermore, we design an End-to-End Underwater Prompt Generator (EUPG), which automatically generates underwater prompts instead of providing explicit foreground points or boxes, enabling accurate and efficient segmentation. Experimental results demonstrate that UWSAM significantly outperforms existing state-of-the-art methods on multiple underwater instance datasets. The dataset and code are available at https://github.com/LiamLian0727/UIIS10K .
Contribute to underwater image analysis research by providing a large-scale underwater object instance segmentation dataset, UIIS10K.
◦
We propose a UWSAM model that improves the efficiency of SAM and demonstrates the possibility of real-time object segmentation in underwater environments.
◦
Enhanced user convenience and improved accuracy through EUPG, an automatic underwater prompt generation technique.
◦
Experimentally verified performance improvement compared to existing state-of-the-art methods.
•
Limitations:
◦
The number of categories in the UIIS10K dataset may be limited to 10.
◦
Further research is needed on generalization performance, focusing only on specific aquatic environments.
◦
Additional analysis and interpretation may be required to improve the performance of the UWSAM model.
◦
Further validation is needed to determine whether EUPG's performance remains consistent across diverse underwater environments.
