Identification of Fish Species and Grading in Fish Market Using Transfer Learning

The identification and recognition of fish species plays a key role in preserving the ecosystem and the fishery industry in quantification on board fishing vessels. Moreover, monitoring methods based on inspection are not effective and take a lot of time. In this paper, we have used a deep learning method to identify six different species of fishes. We followed three major steps: first data augmentation by resizing the images into different sizes. Secondly, make a bounding box and prepare it for the training with YOLOv4 architecture. Finally, a transfer learning method is applied to train the yolov4 model. The data is collected from Nagapattinam fish market in which the model used 80% of the data for training and 20% for testing. According to the experimental result, the proposed model achieves 99.86% accuracy on the test dataset.