{"id":"nvlabs-neuralangelo","name":"neuralangelo","af_score":39.8,"security_score":41.8,"reliability_score":35.0,"what_it_does":"Neuralangelo is an official implementation for high-fidelity neural surface reconstruction, built on the NVIDIA Imaginaire library. It provides training and inference (including isosurface/mesh extraction) for reconstructing a 3D surface from images/video with known camera poses.","best_when":"You have a GPU environment (CUDA) and can follow the repo’s training/data-prep steps to run the pipeline end-to-end for research use.","avoid_when":"You need a lightweight, dependency-minimized library, or you need an external API/service with strong operational guarantees (auth, rate limits, SLA) rather than offline training scripts.","last_evaluated":"2026-03-29T14:59:50.091262+00:00","has_mcp":false,"has_api":false,"auth_methods":[],"has_free_tier":false,"known_gotchas":["No programmatic API surface is provided; integration is via CLI scripts (train.py, extract_mesh.py) and config files.","Pipeline likely depends on substantial GPU memory (README mentions ~24GB for default configuration).","Reproducibility depends on external dependencies (Imaginaire) and correct dataset preprocessing/camera pose format.","Optional W&B logging introduces external service configuration if enabled."],"error_quality":0.0}