Study of Self-prompting Fish Segmentation Method based on SAM2 and YOLO11

Abstract

The 30by30 target was committed at the G7 summit in 2021. This target was committed to conserve more than 30% of each of terrestrial and marine areas as natural environmental areas by 2030. Especially in marine areas, observation and analysis of the underwater environment is necessary for the conservation of natural environmental areas. In this paper, we propose an underwater measurement system developed with a waterproofed omnidirectional camera, along with a system to analyze the captured data. In the proposed analysis system, we analyze fish using a neural network and utilize this analysis information as prompts to perform fish segmentation. Furthermore, we verified the effectiveness of our proposed method using data measured in a real-world environment.