search

nikogamulin / torch-points3d

Pytorch framework for doing deep learning on point clouds.
https://torch-points3d.readthedocs.io/en/latest/
Other
0 stars 0 forks source link

how transformations are applied #2

Open ttsesm opened 3 years ago

ttsesm commented 3 years ago

@nikogamulin I've seen that you are specifying some transformations to be applied in the config file https://github.com/nikogamulin/torch-points3d/blob/6dcd4dffabe7a8d2ca6dcea92b91838124a80b8c/torch_points3d/datasets/regression/superquadrics.py#L210

However, if I check on the corresponding class https://github.com/nikogamulin/torch-points3d/blob/6dcd4dffabe7a8d2ca6dcea92b91838124a80b8c/torch_points3d/datasets/regression/superquadrics.py#L129 I can see that the transformations are not really used. Do you have any idea how to apply the corresponding transformations. I am trying to use for example the GridSampling transform but I am not sure how to apply it.

Thanks.

nikogamulin commented 3 years ago

SuperQuadricsRegressionShapeDataset is a child of BaseDataset and there is a logic for transformations.

ttsesm commented 3 years ago

In my case when I apply these transformations it doesn't seem to be working. I mean for example when you use the Sparseconv3d model it needs to have the coords object which is possible that you can get from the Gridsampling transformation. If I add this transformation into the config file it does not produce this coords object (which currently I am producing manually).