Shrimp recognition using ShrimpNet based on convolutional neural network

Shrimp is a world's important trade goods with high economic value and also one of the most important sources of animal protein. Considering the costs of calculation and hardware, this paper presents a convolutional neural network (CNN) architecture (named as ShrimpNet) to obtain shrimp recognition. The proposed ShrimpNet is an important part of the intelligent shrimp aquaculture which is great helpful for the shrimp aquaculture. The proposed ShrimpNet includes two CNN layers and two fully-connected layers. The collected data set includes six different categories of shrimp that are used to train and test the performance of proposed ShrimpNet. Experimental results show that the proposed ShrimpNet has 95.48% accuracy in shrimp recognition. Therefore, the proposed ShrimpNet is a useful tool with good performance for shrimp recognition.