Object detection is one of the sub-component of computer vision. With recent development in the deep neural networks, many day-to-day problems can be solved. One of the practical problems faced by people is the difficulties in identifying the fish species correctly. Even though there are few studies to solve this problem, those implemented solutions are not easily accessible. The main objective of this study is to implement a mobile application based on deep learning that can detect the fish species and provide information on vitamins and recipes. For this study top selling 16 Sri Lankan fish species are used. In this study, we were able to obtain an accuracy of 77% using YOLO, a convolutional neural network approach. The mobile application takes 12-20 seconds to detect the fish species based on the internet speed.
Supervisor(s): Dr TGI Fernando