Weakly supervised learning using proportion-based information: An application to fisheries acoustics

This paper addresses the inference of probabilistic classification models using weakly supervised learning. In contrast to previous work, the use of proportion-based training data is investigated in combination to non-linear claassification models. An application to fisheries acoustics and fish school classification is considered and experiments are reported for synthetic and real datasets. 1 Problem statement and related work Sonar echosounder mounted on oceanographic and fishing vessels provides a remote sensing device for observing and analyzing the spatio-temporal variations of marine ecosystems [6]. For instance, fish stock assessment for pelagic species such as anchovy or herring are carried out from such acoustic surveys in the Bay of Biscay or the North Sea []. Isolated fish, fish schools or other kinds of aggregation such as plankton layers are viewed by echo sounder as areas depicting a significant echo compared to the background (i.e., water). Besides, the responses of diff...

• Type: Conference
• Year: 2008
• Where: ICPR
• Authors: Carla Scalarin, Jacques Masse, Jean-Marc Boucher, Paul Cauchy, Riwal Lefort, Ronan Fablet
• Keywords: Computer Vision, Fish School Classification, ICPR 2008, Supervised Learning Methods, Trawl Catches