Scene Lighting

Scene lighting research focuses on accurately modeling and manipulating light in images and videos, aiming for realistic rendering and improved perception in computer vision applications. Current efforts leverage deep learning, employing neural networks like transformers and unrolled optimization networks to address challenges such as low-light enhancement, dynamic scene rendering, and physically-based lighting editing from single images. These advancements are crucial for enhancing the realism of virtual environments, improving the robustness of robot perception systems, and enabling novel image and video editing capabilities.