Traditional rendering techniques struggle with optimizing scene parameters due to their non-differentiable nature. This study introduces a novel approach to scene reconstruction using a differentiable ray tracing framework, enabling gradient-based optimization for refining object properties, lighting, and camera parameters. By leveraging automatic differentiation, the proposed method efficiently reconstructs complex scenes with improved accuracy over conventional optimization techniques. Experimental results demonstrate the effectiveness of this approach in minimizing reconstruction error while maintaining computational efficiency. These findings contribute to the advancement of neural rendering and inverse graphics applications.
Akhil Veluru (Thu,) studied this question.