GT Rain

GT-Rain is a research initiative focused on improving single image deraining techniques using a novel real-world dataset of rainy and post-rain images. Current research emphasizes developing robust deep learning models, including transformer-based networks and U-Net architectures, to accurately remove rain artifacts while preserving image detail and handling challenging scenarios like heavy rain and fog. This work is significant for advancing computer vision capabilities in adverse weather conditions, with applications in autonomous driving and other areas requiring reliable image processing in real-world environments. The challenge also highlights the importance of addressing label shift and achieving robust generalization in challenging weather conditions.