POLAR package
Project description
POint cloud LAtent Registration
Getting started
POLAR is a method to simultaneously register numerous highly degraded point clouds corresponding to views of the same unknown reference object.
Installation
[!WARNING] PyTorch3D is required by POLAR. Since its installation is quite specific, you need to install it manually.
If you want to user POLAR with the provided pretrained autoencoder, run
pip install polaregistration
If you want to retrain the autoencoder on your own, run
pip install polaregistration[train]
If you want to use the interactive visualization function, run
pip install polaregistration[vis]
Finally, to install everything, run
pip install polaregistration[all]
Minimal example
from polar import load_sample_data, POLAR
X, degradations, R_abs_gt = load_sample_data()
model = POLAR(**degradations)
X_hat = model.fit_transform(X)
POLAR partially respects the Scikit-Learn Estimator API. Specifically, it has the two main methods:
fit(X: Tensor | Sequence[Tensor]) -> None
fit_transform(X: Tensor | Sequence[Tensor]) -> Tensor | Sequence[Tensor]
They take a list of PyTorch tensors (each of shape $(K_i, 3)$ if point clouds are of varying lengths) or a single batch tensor of shape $(N, K, 3)$ containing all the (same length) views.
Documentation
A documentation, containing full api reference as well as small showcases is available here.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for polaregistration-0.1.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 4004508d98f845e372ef447a3b5bb9d6636d0a04312735d6b89b6d3d08de9241 |
|
MD5 | 31fde39e8ea18eb7e6e172803181a5df |
|
BLAKE2b-256 | 4d5614a240220f8d6f24e54c80e6bbbc0391bc51bf81a1f99a8b5f0373cbe4ff |