Sarr J. M. A. auteur aut IRD Brochier Timothée auteur aut IRD Brehmer Patrice auteur aut IRD Perrot Yannick auteur aut IRD Bah A. auteur aut IRD Sarre A. auteur aut IRD Jeyid M. A. auteur aut IRD Sidibeh M. auteur aut IRD El Ayoubi S. auteur aut IRD text journalArticle eng text/pdf born digital access Quantitative and qualitative analysis of acoustic backscattered signals from the seabed bottom to the sea surface is used worldwide for fish stocks assessment and marine ecosystem monitoring. Huge amounts of raw data are collected yet require tedious expert labeling. This paper focuses on a case study where the ground truth labels are non-obvious: echograms labeling, which is time-consuming and critical for the quality of fisheries and ecological analysis. We investigate how these tasks can benefit from supervised learning algorithms and demonstrate that convolutional neural networks trained with non-stationary datasets can be used to stress parts of a new dataset needing human expert correction. Further development of this approach paves the way toward a standardization of the labeling process in fisheries acoustics and is a good case study for non-obvious data labeling processes. specialized Fisheries acoustics Machine learning Neural network Active acoustics Labeling process Bottom correction 040 122 020 109 113-125 2021 0019-0578 https://www.documentation.ird.fr/hor/fdi:010081020 10.1016/j.isatra.2020.09.018 0019-0578 [F B010081020] https://www.documentation.ird.fr/hor/fdi:010081020 https://www.documentation.ird.fr/intranet/publi/2021/03/010081020.pdf Accès réservé (Intranet de l'IRD) IRD - Base Horizon / Pleins textes 2021-04-01 2025-02-24 fdi:010081020 fre