alexklwong / learning-topology-synthetic-data Goto Github PK

I would like to ask in setup/setup_dataset_vkitti.py. How do we find the correct matching of the vkitti dataset and the kitti dataset? The vkitti does not have a sparse point image, how can I find the corresponding sparse point image in the kitti dataset matching the ground truth depth in vkitti?
Or I just take any kitti dataset so I can create a validity map from kitti?

Thank you.

I got an error when training Scaffnet using virtual KITTI dataset.
Is there a bug using this dataset? Thank you.

tensorflow.python.framework.errors_impl.InvalidArgumentError: indices[2,288,0] = -2147483648 is not in [0, 256) [[{{node scaffnet/GatherV2_3}}]]

Could you provide the running time for evaluation of KITTI dataset in Scaffnet and hardware used?
I would like to see if I could get a higher FPS.
Thank you.

Hello, thank you for sharing your nice works.

It seems that you've also conducted on VOID 0.05% density dataset in arxiv paper.
However, the provided pre-trained weight of ScaffNet was trained on SceneNet 0.50% density, is it right?
If you get a chance, (and if the file still exists) could you upload the ScaffNet's pre-trained weight on 0.05% density in SceneNet?

Best regards,
Jinwoo Jeon