Neural Relighting

Neural relighting aims to realistically change the illumination of images or videos after capture, offering control over lighting effects not possible during original shooting. Current research focuses on developing efficient neural networks, often employing physics-inspired features or compact reflectance representations, to achieve high-quality relighting across diverse objects, poses, and identities, even with limited training data. This field is significant for its potential applications in areas such as augmented reality, virtual production, and digital art, enabling more realistic and controllable image manipulation. Recent advancements leverage techniques like teacher-student frameworks and novel light stage data acquisition methods to improve both accuracy and efficiency.