Complex data labeling with deep learning methods: Lessons from fisheries acoustics
This addresses the tedious expert labeling bottleneck in fisheries acoustics for marine ecosystem monitoring, though it appears incremental as an application of existing methods to this domain.
The paper tackles the problem of labeling non-obvious echograms in fisheries acoustics, which is time-consuming and critical for analysis, by demonstrating that convolutional neural networks trained on non-stationary datasets can identify parts of new datasets requiring human expert correction.
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.