Synthesis of Dimensionality: a Distinctive Approach for High-Fidelity 3D Surface Prediction

EasyChair Preprint 11763

Abstract

This paper introduces an implementation of Neural Radiance Fields (NeRF) for 3D surface prediction. NeRF is a powerful approach for synthesizing complex 3D scenes from sparse 2D observations. In this work, we present a concise neural network architecture for NeRF and utilize synthetic 3D data to train the model. The training process involves optimizing the model parameters to minimize the mean squared error loss between predicted and ground truth surfaces. The results showcase the model's ability to accurately predict 3D surfaces, as demonstrated through visualizations of both ground truth and predicted surfaces. The simplicity of our implementation serves as an accessible entry point for researchers and practitioners interested in exploring NeRF and its applications in 3D surface prediction.

Keyphrases: 3D Surface Prediction, Neural Radiance Fields, machine learning, neural networks, synthetic data

@booklet{EasyChair:11763,
  author    = {Susmita Ghosh and Abhiroop Chatterjee},
  title     = {Synthesis of Dimensionality: a Distinctive Approach for High-Fidelity 3D Surface Prediction},
  howpublished = {EasyChair Preprint 11763},
  year      = {EasyChair, 2024}}