Aquaculture

Fish farming is one of the fastest-growing food systems in the world, but most farms overfeed their fish, wasting money and introducing excess waste into already fragile aquatic environments. The Aquaculture team uses computer vision and behavioral tracking software to monitor fish in real time, identifying hunger cues so farms can feed precisely rather than by guesswork. This semester, the team established a live monitoring system in a Cornell lab and began developing an automated feeder that translates their findings into a practical, deployable tool for fish farmers.

Project Overview

Stakeholders Dr. Eugene Won's lab (tilapia research partner); Kenneth Post Laboratory; future prospective fish farms in Ithaca

Disciplines / Majors Biology, Computer Science, Mechanical Engineering

Team Overview The Aquaculture team works within CUSD to apply machine learning and automation to fish farming systems, with the goal of optimizing feed delivery in ways that benefit both farmers and the surrounding ecosystem. The team is made up of undergraduate students across majors, organized into biological and computational subteams. They work in Kenneth Post Laboratory with tilapia from Dr. Eugene Won's lab.

Problem Statement Traditional aquaculture systems prioritize maximum output without accounting for feed overflow, which wastes money for farmers and generates excess waste and bacterial growth in enclosed water systems. No widely available tool exists to track fish hunger cues or behavioral patterns in real time.

Approach The team uses computer vision and machine learning to recognize fish movement patterns and generate trackable behavioral data. The biological subteam researches prior tracking methods and identifies similar projects, while the computational subteam builds and refines the algorithm and automation pipelines. A permanent remote monitoring setup has been established in Kenneth Post Laboratory.

Key Accomplishments This Semester Developed and refined an algorithm that recognizes fish movement patterns and generates quantitative data on behavior. Built automation pipelines to support data collection. Established a permanent remote monitoring setup for tilapia. Began development of an automated feeder program aligned with observed behavioral patterns.

Next Steps Continue gathering behavioral and hunger-cue data to build out quantitative reports. Develop and test the automated feeder prototype. Pursue a partnership with local fish farms in Ithaca to study different environments. Compile findings toward publishing a paper.

Risks & How They Were Addressed The team operates in a niche, under-resourced field with limited precedent. Ensuring code accuracy and pipeline reliability has required ongoing debugging. The biological-computational integration is complex, but the dual-subteam structure and shared work on the automated feeder have helped bridge both sides.