Artificial intelligence for fish behavior recognition may unlock fishing gear selectivity

Through the advancement of observation systems, our vision has far extended its reach into world fishes, and how they interact with fishing gears—breaking through physical boundaries visually adapting to challenging conditions in marine environments. As sciences step era artificial intelligence (AI), deep learning models now provide tools for researchers process a large amount imagery data (i.e., image sequence, video) on fish behavior more time-efficient cost-effective manner. The latest AI detect categorize species are reaching human-like accuracy. Nevertheless, robust track movements situ under development primarily focused tropical species. Data accurately interpret interactions gears is still lacking, especially temperate fishes. At same time, this an essential selectivity studies advance integrate methods assessing effectiveness modified gears. We here conduct bibliometric analysis review recent advances applications automated tracking, classification, recognition, highlighting may ultimately help improve gear selectivity. further show transforming external stimuli that influence behavior, such as sensory cues background, interpretable features learn distinguish remains challenging. By presenting applied improvements (e.g., Long Short-Term Memory (LSTM), Generative Adversarial Network (GAN), coupled networks), we discuss advances, potential limits meet demands policies sustainable goals, scientists developers continue collaborate building database needed train models.