View Synthesis

View synthesis is the generation of novel images of a scene from new viewpoints, given a set of input views. It's a core problem in 3D reconstruction, virtual reality, and photorealistic rendering.

Neural Radiance Fields (NeRFs) revolutionized this field by using deep learning to encode a scene's volumetric properties. They learn not just surfaces, but how light interacts through space—enabling scenes to be rendered from arbitrary angles.

View synthesis is where geometry, graphics, and machine learning collide—making the leap from pixels to light fields.