Ahmedabad: With AI-enabled tools now widely used to enhance photographs, experts at IIT Gandhinagar have highlighted limitations in supervised learning-based training methods commonly used for image restoration. Ahead of World Photography Day on Aug 19, they said a new image restoration method has yielded better results.A joint team of researchers from IIT-Gn and IIT-BHU has developed Generative Latent Inversion for Blind Face Restoration, or GenR, an AI-powered face restoration framework that does not require paired training data.“It uses a powerful face-generation model to imagine possible clean versions of a damaged image and gradually refines them until the generated face matches the available visual evidence,” said a researcher.According to the researchers, the approach can help enhance CCTV footage, archival photographs and severely damaged images. The team evaluated the tool on tasks including image de-noising, converting low-resolution images into sharper versions, filling missing parts, restoring uncovered portions, and addressing blur and pixelation.
