Skip to main content
ARS Home » Southeast Area » Stoneville, Mississippi » Warmwater Aquaculture Research Unit » Research » Research Project #445930

Research Project: Quanti-Fry: Intelligent Monitoring Framework for Quantification of Fish Eggs and Fry

Location: Warmwater Aquaculture Research Unit

Project Number: 6066-31000-016-042-S
Project Type: Non-Assistance Cooperative Agreement

Start Date: Mar 18, 2024
End Date: Mar 17, 2025

Objective:
To streamline the quantification of fish eggs and fry through automation. We propose leveraging Artificial Intelligence (AI)--based computer vision (CV) methods to autonomously compare historical data of fish eggs and fry, creating an efficient approach for enumerating egg and fish samples. This process will generate a unique identification profile facilitating the quantification of fish eggs and fry, with future potential for fish tagging to monitor their behavior. The immediate outcomes will include a cost-effective AI-enabled monitoring system that quantifies each collected sample. The proposed autonomous camera system will not only enumerate but also estimate the mass of eggs and fry. In the long term, these methods and algorithms can be scaled to other fish species, enabling broad adoption of this technology across regions and industries.

Approach:
A camera system will be designed that incorporates an onboard processor for object detection, enabling the counting of eggs and fry from images captured by the smart camera. To enhance accuracy, we intend to develop algorithms measuring the body mass of fish fry. These algorithms will integrate metadata collected from the samples to autonomously estimate fish volume, estimate sample weights, and estimate the number of fry in each sample. This task will create an extensive database of catfish eggs and fry photographs. Images of catfish from eggs, sac fry, and swim-up fry will be organized, structured, and annotated per the information given by the USDA team. The algorithm will be deployed in the onboard processor integrated into the camera system. Through this automated identification process, we will explore methods to calculate the weight and enumerate the catfish through our customized computer-vision modeling techniques. The database will include images of eggs, sac fry, and swim-up fry of channel catfish.