neuralangelo

Why we included this project

Neuralangelo is the official implementation of the CVPR 2023 paper that reconstructs dense, high-fidelity 3D surfaces from posed video. It captures fine geometry like brick textures and stone carvings that many image-based neural rendering methods blur over. The repo includes configuration files, prebuilt Docker images, and a documented COLMAP preprocessing pipeline, so a reproducible baseline is realistic. But it's a research training pipeline: you need known camera poses and GPU time, not a turnkey app. That makes it a solid starting point for researchers and engineers who want to build their own reconstruction experiments on a serious, well-documented codebase.