Inverse Rendering

Inverse rendering aims to reconstruct a 3D scene's geometry, materials, and lighting from 2D images, enabling realistic relighting and novel view synthesis. Current research heavily utilizes neural networks, particularly radiance fields (NeRFs) and Gaussian splatting, often incorporating physically-based rendering models and techniques like differentiable rendering and Monte Carlo sampling to improve accuracy and efficiency. This field is significant for advancing computer graphics, robotics, and virtual/augmented reality applications by enabling more realistic scene modeling and manipulation from readily available image data. Furthermore, the development of new datasets and improved algorithms are driving progress in handling complex materials like glossy surfaces and accurately modeling indirect lighting effects.