A Comparative Study of Machine Learning for iOS based on Siam Betta Mobile Application

Siam Betta, or Siamese fighting fish, is a popular fish in the a quarium trade. However, due to its variety, species classification is an obstacle f or many amateur aquatic animal breeders. Thus, to facilitate those amateurs on Thai Betta' species prediction, in this paper, some machine learning techniques have been applied, studied and deployed in an iOS application. Apple's Create ML framework and Microsoft's Custom Vision framework are investigated and used to build the machine learning model (mlmodel). The mlmodel is built from 6 species of Thai Betta images. In this paper, the performance of both frameworks in terms of the step of the training dataset, the accuracy parameters and the processing time by varying the number of images dataset from 350 to 1400 images have been compared. The results show that the Create ML framework predicted 100% correct at the number of images dataset equal and higher than 875 images but it takes the average processing time higher than that of the training model with Custom Vision around 2.1686 times at 50 images dataset in each class, and 4.4924 times at 200 images dataset in each class.