Fish Inspection Systems using Parallele Neural Network Chips and an Image Knowledge Builder.

Anne Menendez, Guy Paillet

A generic image learning system, CogniSight, is being used for inspecting fishes before filleting. More than thirty systems have been deployed on seven fishing vessels in Norway and Iceland. Each CogniSight is using four hardware neural networks chips (312 neurons) based on natively parallel hardwired architecture performing real time training and non-linear classification (RBF). These systems are trained and the learning can be reinforced directly by the ship crew using Image Knowledge Builder, show and tell training and validation software. Most of the systems have been deployed for three years or more. They reduce significantly the number of crew (up to six persons) and are reducing the time at sea by up to 15%. These systems brought a strong return of the investment to the fishing fleet and, in turn, increased significantly the market shares of Pisces Industries, the company manufacturing the filleting machines.

Subjects: 19.1 Perception; 14. Neural Networks

Submitted: Apr 3, 2007


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.