This project involves designing and training a Neural Radiance Field (NeRF) to render 3D scenes. NeRF encodes scene information using a Multilayer Perceptron (fully connected neural network), enabling realistic rendering from different viewpoints. (Source)
Figure 1: Output after 1000 iterations
After 2000 iterations with 64 depth samples per ray. The training utilized GPU acceleration and processed camera parameters, images, and scene details to render 3D scenes.
Figure 2: Rendered Scene using NeRF
Georgia Tech (Fall 2024)