<![CDATA[This research was conducted to optimize the use of the convolutional neural network (CNN) method in classification of krasak river sand and white river sand in the Muntilan area, Central Java with the aim of knowing or recognizing two types of sand based on digital images, the number of datasets collected amounted to 1,305 consisting of 697 krasak river sand image data and 608 white river sand image data, which then after going through Data Pre-processing process obtained 1200 image data that could be used in this study, each consisting of 600 data from each type of sand. This study uses 5 different input image sizes namely 32x32x3, 60x60x3, 122x122x3, 175x175x3, 225x225x3, to find out which input image size can produce the most effective accuracy data, the results found that the 122x122x3 input image size with epoch 10 and 3 layers CNN architecture is the best input image size with an accuracy of 99, 58% in recognizing and classifying Krasak river sand and white river sand, which then results of the most accurate input image were implemented in form of a Graphical User Interface (GUI) in facilitating the recognition of the classification of Krasak river sand and white river sand.]]>
Copyrights © 2024
