Testing The Limits Of AI: Deep Learning Goes Deep Sea Fishing


Deep learning has been making waves in many industries. Now, the fishing industry also wants to incorporate it to improve the quality and quantity of the catch.

This is the goal of Nature Conservancy, a non-profit organization that works with tuna fishing companies in the Pacific Islands. Through deep learning technology, they hope to help these companies identify fish and count their catch more easily. Most of all, they hope that through the use of cutting-edge technology, it will protect those animals that need not be caught.

According to them, a lot of these animals, especially protected sea animals like sea turtles and sharks, get accidentally caught by fishermen. Artificial technology will help fishermen avoid this by helping them identify fish faces. As Nature Conservancy project director Mark Zimring said, it will help them tell the difference between a turtle and a tuna, and "flag when a shark is on board."

At present, fish identification is a slow process because there's no technology to do that. A ten-hour fishing trip will take six hours to review without the proper tools. Furthermore, these fishing boats do not really have some expert that would assure all fish are accounted for. One fishing company has but the person only accompanies them 10 times out of the 200 trips they make.

To do this, the company has teamed up with Kaggle to develop an algorithm that would help with the identification process. This algorithm will be embedded in a fully automated software tool which fishermen can use. Nature Conservancy has provided the startup with hundreds of fishing footage which were then sliced into thousands of images. They will use these images to create the algorithm.

This is going to be a difficult challenge because of the obstacles including poor lighting conditions and rain which obstructs camera view. However, they want to test the limits of artificial intelligence and hope that it will be a success.

© 2024 University Herald, All rights reserved. Do not reproduce without permission.
Join the Discussion
Real Time Analytics