Pyramid NeRF: Frequency Guided Fast Radiance Field Optimization

Check out the paper: https://link.springer.com/article/10.1007/s11263-023-01829-3

This repo contains the code for Pyramid NeRF, more installation details refer to KiloNeRF: https://github.com/creiser/kilonerf

Prerequisites

• OS: Ubuntu 16.04.1 LTS
• GPU: >= NVIDIA GTX 2080 Ti
• Python package manager conda

Data

To generate multiscale image datasets run: python generate_data.py Note that camera intrinsics and extrinsics can be copied from the original dataset, f, cx and cy decrease proportionally.

Usage

To benchmark a trained model run:
bash benchmark.sh

You can launch the interactive viewer by running:
bash render_to_screen.sh

To train a model yourself run
bash Chair_train.sh

The default dataset is Synthetic_NeRF_Chair, you can adjust the dataset by setting the dataset variable in the respective script.