Detection of Resting Zones Based on PSPNet in Salmon Scale Images

Sustainable fisheries resource management is essential due to the increasingly poor catch of chum salmon in Japan. This paper presents a neural network-based method for detecting resting zones in salmon scale images. Specifically, the proposed method uses a semantic segmentation model that is trained to effectively recognize scales with different shapes and patterns. Automation of the fast detection of resting zones can increase the utilization of the database of migrating chum salmon and contribute to our analysis of poor catches.