kwea123/ngp_pl
A PyTorch implementation of Instant-NGP for neural radiance field (NeRF) 3D reconstruction and novel view synthesis.

Velocity · 7d
+0.9
★ / day
Trend
→steady
star history
This repository implements the Instant Neural Graphics Primitives (Instant-NGP) algorithm for Neural Radiance Fields (NeRF) using PyTorch and CUDA. It leverages PyTorch Lightning to train neural networks that represent 3D scenes and generate novel viewpoints from sparse input images. The project provides a concise interface for research purposes, achieving high-quality 3D reconstructions at high speed.