<![CDATA[Artificial Neural Networks are known to provide a good model forclassification. The goal of this research is to classify books in Bahasa (Bahasa Indonesia) using its cover. The data is in the form of scanned images, each with the size of 300 cm height, 130 cm width, and 96 dpi image resolution the research conducted features extraction using image processing method, MSER (Maximally Stable Externally Regions) to identify the area of book title, and Tesseract Optical Character Recognition (OCR) to detect the title. Next, features extracted from MSER and OCR are converted into a numerical matrix as the input to the Backpropagation Artificial Neural Network. The accuracy obtained using one hidden layer and 15 neurons is 63.31%. Meanwhile, the evaluation using 2 hidden layers with a combination of 15 and 35 neurons resulted in accuracy of 79.89%. The ability of the model to classify the book was affected by the image quality, variation, and number of training data.]]>
Copyrights © 2020
