<![CDATA[Abstract painting always has its own place for the fans, the irregular shape in it, and the emotions depicted in the painting, make many people amazed to see it. The success of this abstract image sparked the idea of being able to create an abstract image using Deep Learning Technology. Generative Adversarial Networks (GANs) is one of the Deep Learning technologies that can create it. With the GANs method which has Generator and Discriminator functions in it, it is possible for someone to be able to create it. The generator functions to generate new data through training the data(train), and the Discriminator functions to determine whether the new data is fake or not data through training (train) comparing the generator results with the original data. These two functions are used to create abstract images. Abstract images were obtained through training in 1369 paintings of nature, landscapes, and flowers. The images are trained by comparing the number of epochs used and the results of the abstract images generated from each epoch. The epoch will be divided into three parts, namely the first training using 10 epochs, the second training using 100 epochs, and the third training using 1000 epochs. In this journal, we will compare the results of the three trainings and reach a conclusion which training produces the best abstract image according to the author. From the training, 1000 epoch training was obtained which produces good abstract images.]]>
Copyrights © 2022
