<![CDATA[The challenge of neural networks to process visual art judgments like humans inspired Gatys et al and in 2015 they succeeded in creating neural style transfer (NST) that can transfer European artistic image styles to other images. At present, research related to NST has been widely conducted, but its use with a convolutional autoencoder (CAE) as one of the NN architectures capable of compressing NST output is still rare. This research intends to design an NST system with CAE as an additional architecture in charge of the compression process while maintaining the force transferred. As a substitute for European-style artistic images, batik is used as an original Indonesian artistic work. NST and compression images will be measured using structural similarity index measure (SSIM) evaluation metrics. The evaluation results showed that the system designed managed to get an average SSIM score of 0.67 out of 1 and an average value of storage size reduction ratio of 37.43% from the original size. Then, the survey showed that the quality of the compressed image was quite good with a score of 64.09% and the compressed image was quite usable in the field of work of each respondent with a score of 49.09%.]]>
Copyrights © 2023
