Publications 📑

Evaluating GAN Architectures for Generating Images of Defective Facades

We evaluated the performance of DCGAN, BigGAN, and Lightweight GAN architectures in generating facade images featuring four types of defects: erosion, efflorescence, missing mortar, and spalling. Our primary goal was to generate additional data to augment existing datasets. Among the models tested, Lightweight GAN proved to be particularly effective for this application.