Aqua Vision Deep Learning Research

An Open-Source Deep Learning Project to Identify Trout Species

Aqua Vision Deep Learning Research Parent Template

This project is aimed at accurately identifying trout species from images using computer vision deep learning. The inspiration behind this project comes from spending countless hours trout fishing in the Northwest United States (Montana, Washinton, Wyoming) and talking with professional fishing guides about trout identification.

Currently, there is a dataset of 1500+ hand labeled trout images, a web-app and an API endpoint for making trout predictions from the deep learning model.

The continued goals of the project are to:

  1. Build an Open-Source Trout Photo Library to continue to improve upon the deep learning model accuracy and increase the number of supporting classes for trout species identification (eventually expanding to other fish types)
  2. Label Images with Bounding Boxes to incorporate Object Detection into the Model
  3. Geotag Photos with coordinates and timestamp for future Trout species research (population distributions, migratory patterns, crossbreeding, geospatial analysis, etc)
  4. Build a supporting native mobile app