arxivcs.SDcs.LG2026-07-07
A Self-Supervised Approach for Minimal-Annotation Hydroacoustic Data Exploration
Pierre-Yves Raumer, Axel Marmoret, Dorian Cazau, Anatole Gros-Martial, Richard Dreo, Maelle Torterotot, et al.
Passive hydroacoustic monitoring often generates large volumes of continuous recordings that are only partially exploited due to the cost of manual annotation. Supervised detection methods perform well but require large labeled datasets, seldom available for rare signals or under…