<![CDATA[This study evaluates the impact of variations in the number of layers on the implementation of the Convolutional Neural Network (CNN) algorithm in a color-based object identification and categorization system, using python language supported by the TensorFlow/Keras framework. The data used is a collection of visual data in the form of red and white cups divided into a proportion of 90% training data and 10% testing data in the dataset in this study which amounted to 62 red cup data and 59 white cup data. Testing was carried out by comparing three different convolution layer configurations of 1, 2, and 3 layers, where each configuration was integrated with a max pooling and fully connected layer. The results of the study showed an accuracy of 92%, precision of 93%, recall of 92%, and f1-score of 92%. On the other hand, the application of two and three convolution layers actually showed a significant decline with an accuracy of only 46%.]]>
Copyrights © 2025
