Getting started¶
From an empty environment to a trained INR in two short pages.
-
Requirements, the optional extras, and the benchmark data.
-
Fit FUTON to an image, read the result, and swap in another model.
The package at a glance¶
Everything is importable from the top-level package:
import inro
inro.FUTON, inro.SIREN, inro.InstantNGP # models
inro.ImageCoordinateDataset, inro.OccupancyCoordinateDataset # datasets
inro.train, inro.evaluate, inro.chunked_inference # training and evaluation
inro.psnr, inro.ssim, inro.lpips, inro.iou # metrics
inro.futon_bound # FUTON's error bound
inro.nerf # radiance fields