I had the opportunity to collaborate on a project aimed at restoring damaged murals, contributing to the preservation of cultural heritage. We conducted a detailed comparative analysis of three cutting-edge inpainting models: AOT-GAN, DeepFillv2, and PConv. After thorough evaluation, we developed an innovative solution using a weighted average approach to combine the strengths of all three models for optimal reconstruction results.Technically, I leveraged Python and Torch for model development and training, and utilized PIL for effective image processing.