PERBANDINGAN OPTIMIZER, BATCH SIZE DAN EPOCH PADA METODE CONVOLUTION NEURAL NETWORK

Abstract

Buffalo meat and beef are two types of red meat that are widely consumed by the public. The demand for meat increases every year. However, not all types of meat can be eaten by Indonesians, such as pork, so the price of pork in Indonesia is lower than the price of beef and buffalo. In general, the texture and colour of pork, beef and buffalo are almost the same. In the introduction of meat, it is only done directly from the colour, texture, and fibre of the type of meat. However, meat circulating in the community is often mixed between beef, buffalo meat and pork. Distinguishing beef, buffalo and pork must first recognise the characteristics of each type of meat, because there are limitations to the human sense of sight in distinguishing between them. In the use of technology with the help of digital images to determine the most optimal optimizer, batch size and epoch in meat classification, using the Convolutional Neural Network (CNN) method with NasNetmobil Architecture. The data set used is 3000 images divided into three classes, with a division of 2400 training data images, 300 testing data images, 300 validation data images. The results showed that the Adam optimiser, batch size 62 and epoch 20 produced an accuracy of 99.00% and a loss value of 0.0243. Keywords: Convolutional Neural Network, Buffalo and Beef Classification,