Conditional GAN Approach for Image to Image Translation on Potato Leaf

Abstract

In taking pictures using a camera, for example, it often results in poor image quality. The resulting image quality can be in the form of black and white images. This can interfere with further processing of the image. Specifically, this can happen in the agricultural sector, such as the results of a bad potato leaf image (black and white). In fact, the images of leaves from agricultural land will be used for image processing which can help with the planting process, for example. This is of course very disturbing, so a way is needed to translate a gray scale image into an image that has a color resembling the shape of an actual leaf. In fact, the development of Deep Learning with various models has grown rapidly, one of which is the Conditional GAN which can process image to image translation. Seeing the purpose of the Conditional GAN, this paper implements the Pix2Pix model based on the Conditional GAN which aims to translate black and white images into images that have color. This experiment produces images with colors that match the original image with good quality. So that with the results of this experiment, it is hoped that problems with taking images that are sometimes not good enough can be solved by image translation using deep learning.