Classification of Dog and Cat Images using the CNN Method

Abstract

Blind people can be defined as those people who are unable to see objects or pictures around them with their eyes. This inability becomes an issue for them when dealing with objects or images in front of them. These problems lead to the novelty of this study that is to recognize objects or images around blind people with the CNN algorithm. Dogs and cats were used as objects in this study. These object recognitions used Deep Learning, a relatively new science in the field of machine learning. Deep learning works like the human brain's ability to recognize an object. In this study, the objects that were used were pictures of a dog and a cat. This study used 3 types of data, namely training, validation, and testing data. The data training consisted of dog data with a total of 1000 images and cat data with a total of 1000 images. Data validation consisted of 500 dog data  and 500 cat data. The CCN architecture employed 3 convolution layers. The layer was convolution 1 using 16 filters of kernel size 3x3, the second convolution using 32 filters of  kernel size 3x3 and the third using 64 filters of kernel size 3x3. While the data testing consisted of 51dog data and 27 cat data. The method used to analyze the image was CNN. The input was an image with a size of 150x150 pixels with 3 channels, namely R, G, and B. This classification went through a performance test with the Confusion Matrix and it obtained 45% precision, 45% recall and 45% f1-score. From these results it can be concluded that the accuracy values should be improved.