-
Notifications
You must be signed in to change notification settings - Fork 188
New issue
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
Voxelized data is shifted upwards #49
Comments
Good catch. I already merged your changes and will regenerate the data within the next days. Luckily, the error did not invalidate the results from the metrics. I will write here, when I changed the data and everything is ready for download. Thanks again for finding this problem and notifying us about it. |
You're welcome. Thanks for the nice work with the dataset. |
Did you already find the time to fix the data on the server? I'm unsure whether to submit with the 'old/shifted' voxel data or the 'new/fixed' data. |
As always, different other things needed my attention. I hope to find soon time to look at the issue. Sorry for the delay :( |
Hi @risteon, I finally managed to take the time to regenerate the data. I visually verified that the data looks okay. I generate the data now with the correct counter. Hope that resolves the issue then. You can now download the new data at http://semantic-kitti.org/dataset.html#download. Bonus: I also now generate all scans for the training data which people now download. |
Hi @jbehley, |
Dear authors of SemanticKITTI,
I've made the following observation:
The voxelized annotated point clouds (input grid, target grid, ...) are shifted one voxel upward due to a bug in the voxelizer
saveVoxelGrid
routine. This gets apparent when plotting the input point cloud together with the voxelized data.See this pull request for the SemanticKITTI voxelizer:
jbehley/voxelizer#4
The text was updated successfully, but these errors were encountered: