search

nianticlabs / ace

[CVPR 2023 - Highlight] Accelerated Coordinate Encoding (ACE): Learning to Relocalize in Minutes using RGB and Poses
https://nianticlabs.github.io/ace
Other
350 stars 34 forks source link

Windows #1

Closed Tobe2d closed 1 year ago

Tobe2d commented 1 year ago

Any chance for windows users?

tcavallari commented 1 year ago

Hello, thanks for your question. Unfortunately we have no plans to support Windows.

That said, PyTorch has Windows support, so it should be possible to create a different conda environment that works there.

Anyways, you can start with: 

conda create --name ace python=3.8
conda activate ace
conda install pytorch=2.0.0 torchvision=0.15.0 pytorch-cuda=11.8 -c pytorch -c nvidia
conda install scikit-image
conda install -c conda-forge opencv=4.4
conda install -c conda-forge trimesh
conda install -c conda-forge pyrender
conda install -c anaconda libglu
conda install -c conda-forge matplotlib

and adapt the packages/versions to use variants that are available on Windows.

Building the C++/Python bindings (in the dsacstar folder) might require some work, but we didn't use anything Linux-specific, so they should work on Windows, eventually.

ebrach commented 1 year ago

Regarding the dsacstar bindings: They are not too different from the NG-RANSAC bindings. Someone has successfully ported those to Windows 10 and documented their process here: https://github.com/vislearn/ngransac/issues/21

Some of that will have to be adjusted because the library versions are different (e.g. NG-RANSAC uses an older version of OpenCV). But maybe it can serve as a template.

tcavallari commented 1 year ago

Closing for now.