This repository contains a PyTorch implementation of NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis. NeRF is a groundbreaking method introduced by Mildenhall et al. in their ECCV 2020 paper, which enables the synthesis of novel views of complex 3D scenes.
 |
|---|
NeRF represents scenes as neural radiance fields, enabling high-fidelity 3D scene reconstruction and view synthesis. This implementation leverages PyTorch for efficient computation and training. Whether you're a researcher looking to dive into state-of-the-art neural rendering or a developer interested in 3D graphics and computer vision, this project provides a robust starting point.
Follow these steps to set up your environment, install dependencies, and begin training your own NeRF models.
To configure the Python virtual environment and install dependencies, run the following commands in your shell:
# Clone the repository
git clone https://github.com/shub-garg/NERF-torch.git
cd NERF-torchvirtualenv venv -p python3.8
source venv/bin/activate
pip install -r requirements.txt
You may use alternative tools like conda if preferred. Ensure the correct dependency versions are installed to reproduce the results.
Download the synthetic dataset required for training by running the script below:
sh scripts/download_data.shThis script creates a data directory under the project root (torch-NeRF) and downloads datasets provided by the original NeRF authors.
The configuration is pre-set for the lego scene from the Blender dataset. Start training by running:
python torch_nerf/runners/run_train.pyOnce the scene representation is trained, render it using:
python torch_nerf/runners/run_render.pyUpdate the yaml configuration file under config/train_params to specify the path to the checkpoint file. The rendered images will be saved in the render_out directory.
To compile the rendered frames into a video, use the following script:
python scripts/make_video.py
