Detecting Leftover Food and the Shrimp for Estimating of the Shrimp Body Length Based on CNN

Controlling the body length of the shrimp and estimating leftover food are the major indicators for feeding strategies in shrimp farms. Normally, the works are done by naked-eye observation and mostly counted on the experiences. Nevertheless, it could sometimes be subjective, especially with leftover food. The miscalculation could affect to the feeding plans of farmer, and hence, commercial losses in shrimp aquaculture. In this paper, a new approach to automatically calculate the shrimp body length and to estimate leftover food in pond is given by using a underwater cameras and a convolutional neural network (CNN) model. The proposed method obtained a mAP of 87.3% in the shrimp and leftover food detection and localization. In addition, just 7% of a MSE in the body length of shrimp is produced.