Object Removal

Object removal aims to seamlessly erase unwanted objects from images or videos while maintaining visual coherence. Current research focuses on improving the realism and efficiency of object removal using various approaches, including diffusion models, Gaussian splatting, and radiance fields, often incorporating techniques like task-decoupling, mask optimization, and multi-view consistency checks. These advancements are significant for applications ranging from image editing and 3D scene manipulation to robotics and autonomous driving, where accurate and efficient removal of dynamic objects is crucial for reliable map creation and navigation.