Fish Biomass Estimation Based on Object Detection Using YOLOv7

Fish biomass estimation is an important parameter in aquaculture to measure fish health and identify fish’s feeding patterns and maturity levels. Fisheries conduct regular assessments to estimate biomass, optimize daily feeding behaviors, maintain stocking density, and determine the optimal harvest time. The prevailing assessments involve manual sampling, which is laborious and other methods involve expensive sensors to estimate the biomass. By estimating each fish’s length using calibrated camera images of fish and then calculating the biomass using a calibration curve that expresses the length-mass relation of them, the proposed work seeks to estimate the biomass of fishes in a non-intrusive manner. The system is integrated with Jetson Nano, so the entire module can easily be deployed on a well-lit setup and the biomass estimates are regularly shared with the users. Thus this system helps to calculate the overall biomass of the area so that the growth rates are tracked and feeding formulas are established efficiently.