A Deep Learning Approach to Assist Sustainability of Demersal Trawling Operations

Bycatch in demersal trawl fisheries challenges their sustainability despite the implementation of various gear technical regulations. A step towards extended control over catch process can be established through a real-time monitoring tool that will allow fishers to react unwanted compositions. In this study, for first time commercial fishery sector, we introduce an automated description leverages state-of-the-art region based convolutional neural network (Mask R-CNN) architecture and builds upon in-trawl novel image acquisition system. The system is optimized applications Nephrops enables classification count items during fishing operation. detector robustness was improved with augmentation techniques applied training on custom high-resolution dataset obtained extensive trawling. resulting algorithms were tested video footage representing both normal towing haul-back conditions. algorithm F-score 0.79. compared manual showing low absolute error towing. Current practices are carried out without any indications composition nor whether enters gear. Hence, proposed solution provides substantial contribution making type more targeted, paving way further optimization activities aiming at increasing target while reducing bycatch.