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
359 stars 34 forks source link

generate reconstructions #6

Closed Fungungun closed 1 year ago

Fungungun commented 1 year ago

Hi

Thanks for your work. Just wondering if it's possible to generate the point cloud of the scene with the trained weights as shown in the demo video?

Thanks

ebrach commented 1 year ago

Hi,

yes, that is possible. The visualisation code is fully integrated, and you can create these demo videos yourself. Just run the ACE evaluation script with a pre-trained model, and append --render_visualization True.

If you want to export the point cloud, that is currently not implemented but possible as well. See how the test script extracts the point cloud from the network here by using this function.

The easiest way for an export is to dump the returned pc_xyz, pc_clr tensors to a text file with XYZRGB per point on a new line. You can open these files with CloudCompare or similar tools.

tcavallari commented 1 year ago

Closing for now. Happy to continue the discussion if more questions come up.