This repository provides code for the RCAN model introduced in the ECCV 2018 paper. The model uses a residual in residual (RIR) structure with channel attention mechanisms to form very deep networks for single image super-resolution. The architecture consists of residual groups with long skip connections, where each group contains residual blocks with channel attention to adaptively rescale features. Built on the EDSR PyTorch codebase and tested on standard GPU environments.