View synthesis and 3D scene reconstruction are important topics of computer vision. This project implements the ground-breaking work of Neural Radiance Field (NeRF) based on the original paper from 2020. We implement the core NeRF algorithm, including the fully-connected neural network, volume renderer, as well as key features such as frequency-based positional encoding and hierarchical sampling. We use PyTorch Lightning as our framework and allow the neural network to be trained on the GPU with different memory limitations.