Bags-of-Features for fish school cluster characterization in pelagic ecosystems: application to the discrimination of juvenile and adult anchovy clusters off Peru Ronan Fablet, Paul Gay, Salvador Peraltilla, Cecilia Peña, Ramiro Castillo, Arnaud Bertrand. Fisheries acoustics data presents a great potential for the characterization of pelagic ecosystems, and especially the spatial distribution of pelagic fish. Whereas previous work has mainly focused on the detection, characterization and recognition of individual fish schools, we here addressed the characterization and discrimination of fish school clusters. The proposed scheme relied on the application to acoustic echograms of the Bags-of-Features approach. The latter is widely exploited for pattern recognition issues and relies on extraction and categorization of objects in images; the image descriptor being formed by the count of objects in each object category. This approach is particularly suited to fisheries acoustic data where fish schools appear as a natural and meaningful object concept. We applied this approach to the discrimination of juvenile and adult anchovy clusters off Peru. Echogram-level discrimination performance raised between 88% and 92% of correct classification for different survey datasets. Significant improvement (about 10% of correct classification) was reported compared to previously proposed school-based echogram-level characteristics. The proposed school cluster classification model was applied to the mapping of juvenile and adult anchovy population from routine acoustic survey data