search

chrischoy / SpatioTemporalSegmentation

4D Spatio-Temporal Semantic Segmentation on a 3D video (a sequence of 3D scans)
MIT License
288 stars 51 forks source link

Point-level predictions from voxels #40

Closed amineremache closed 4 years ago

amineremache commented 4 years ago

Hello Chris,

In the paper you are mentioning that to get point-level predictions, you are propagating them from the voxels.

image

And from what I understood, you need to use the argument --evaluate_original_pointcloud, but as you can see: The "pointcloud" is not implemented.

image

So I thought maybe I should use the other flag, --test_original_pointcloud, which uses the save_predictions function.

image

But the function tries to import OnlineVoxelizationDatasetBase , that doesn't exist anywhere in the code ( except where it is being called )

image

image

My question is : Am I doing things the right way ? or is there something that I am missing ?

Thank you

chrischoy commented 4 years ago

Please check out https://github.com/StanfordVL/MinkowskiEngine/blob/master/examples/indoor.py#L126

amineremache commented 4 years ago

Please check out https://github.com/StanfordVL/MinkowskiEngine/blob/master/examples/indoor.py#L126

The example is working, thank you.