# Inverse Rendering

URAvatar is a novel avatar generation technology that enables the creation of realistic, re-lightable head avatars through smartphone scans under unknown lighting conditions. Unlike traditional methods that estimate reflectance parameters through inverse rendering, URAvatar directly simulates learned radiative transfer, effectively integrating global illumination into real-time rendering. This technology is significant as it can reconstruct head models that appear realistic in multiple environments from a single environmental smartphone scan, enabling real-time driving and re-lighting.

DiPIR is a physics-based method co-developed by the Toronto AI Lab and NVIDIA Research, which recovers scene lighting from a single image, allowing virtual objects to be inserted realistically into both indoor and outdoor scenes. This technology not only optimizes material and tone mapping but also automatically adjusts to different environments to enhance the realism of the images.

IntrinsicAnything is an advanced image inverse rendering technology that optimizes the material recovery process through the learning of diffusion models, addressing the issue of object material recovery in images captured under unknown static lighting conditions. This technology learns material priors through generative models, decomposing the rendering equation into diffuse and specular reflection terms, and effectively resolving ambiguity issues in the inverse rendering process by training with a rich dataset of existing 3D objects. Additionally, this technology has developed a coarse-to-fine training strategy, using estimated materials to guide the diffusion model to produce multi-view consistency constraints, thereby achieving more stable and accurate results.