Shadow Generation

Shadow generation and manipulation are active research areas in computer vision, focusing on accurately detecting, removing, and synthesizing shadows in images and videos. Current research employs deep learning models, including neural radiance fields (NeRFs), transformers, and diffusion models, often incorporating physically-based rendering techniques and novel loss functions to improve realism and controllability. These advancements are crucial for enhancing image and video quality, improving the realism of virtual and augmented reality environments, and enabling applications such as object compositing and scene editing. The development of large-scale datasets and new evaluation metrics is also driving progress in this field.