Translation Of Architectural Sketch Images To Building Images Using CycleGAN

Sketching and drawing is the hallmark of an architect's toolkit. Architect needs to somehow show his idea from his sketch and also designing a sketch image from a building image will help the architects to record the different designs. Hence, using image-to-image translation strategies, it is possible to convert architectural vector designs into actual building images and vice versa. Image-to-image translation is becoming more popular as a result of advancements in information technology and computer vision. In a class of vision and graphic problems known as “image-to-image translation”, the objective is to discover the relationship between an output image and an input image. The main objective of the model in this study is to convert architectural sketches into building images and vice versa using Image to image translation on unpaired image datasets to help reduce the communication gap between architects and their clients. A suitable GAN model called Cycle GAN has been used to train on collected unpaired image dataset. Utilizing the trained model, a web application is created that is open for everyone. A Python library called flask is used to create the web application's front end in order to provide a straight-forward and user-friendly interface.