Super-resolution of low-resolution images obtained from BSD200, Urban100, Manga109, Set5, etc. using a GAN-based model, taking ideas from the SRGAN (Ledig, C. et al., 2017) and ESRGAN (Wang, X. et al., 2018) models.
Utilized PyTorch framework for development. Used a NVIDIA GeForce GTX 1070 GPU machine to facilitate training of the model.
Some Training Results
Low-resolution / Original High-resolution / Generated Super-resolution
Some Testing Results
Low-resolution / Original High-resolution / Generated Super-resolution