ANALISIS PENGGUNAAN KOMBINASI WORD EMBEDDING, RESNET, DAN GRU PADA MODEL PIX2CODE
Fadly Triansyah Rahman, Farhan Rahmat Abdillah, Transmissia Semiawan, Nurjannah Syakrani
Abstract
The pix2code model is a machine learning model that can automate the mockup implementation process into code from a Graphical User Interface (GUI). First developed by Tony Beltramelli, this model has a rating of 77%. The development of the pix2code model is considered important because, by increasing accuracy, the code generated by the model is accurate. This study develops a pix2code model using a combination of word embedding, Residual Network (ResNet), and GRU methods. The data used comes from the research of Tony Beltramelli as many as 3500 datasets consisting of mockup images and the context of the image. The results of this study indicate that the combination of word embedding, ResNet, and GRU has succeeded in increasing the accuracy of the pix2code model from 77% to 80%. In addition, the value of the distribution variance of the pix2code model's accuracy decreased from 0.078782 to 0.046656. This shows that with the variation of the data used, the accuracy of the developed pix2code model becomes more stable.