FishNet: A Unified Embedding for Salmon Recognition

  • 2020-10-20 17:35:01
  • Bjørn Magnus Mathisen, Kerstin Bach, Espen Meidell, Håkon Måløy, Edvard Schreiner Sjøblom
  • 0

Abstract

Identifying individual salmon can be very beneficial for the aquacultureindustry as it enables monitoring and analyzing fish behavior and welfare. Foraquaculture researchers identifying individual salmon is imperative to theirresearch. The current methods of individual salmon tagging and tracking rely onphysical interaction with the fish. This process is inefficient and can causephysical harm and stress for the salmon. In this paper we propose FishNet,based on a deep learning technique that has been successfully used foridentifying humans, to identify salmon.We create a dataset of labeled fishimages and then test the performance of the FishNet architecture. Ourexperiments show that this architecture learns a useful representation based onimages of salmon heads. Further, we show that good performance can be achievedwith relatively small neural network models: FishNet achieves a false positiverate of 1\% and a true positive rate of 96\%.

 

Quick Read (beta)

loading the full paper ...