Sequence 21 broken: "Missing prediction files" error when submitting to completion benchmark on CodaLab
See original GitHub issueHi @jbehley,
I’m using the updated scene completion data V1.1 (downloaded September 10th, verified with a fresh download from just now). However, when submitting my prediction files to CodaLab I get the following python exception that points to missing files:
WARNING: Your kernel does not support swap limit capabilities or the cgroup is not mounted. Memory limited without swap.
Traceback (most recent call last):
File "/tmp/codalab/tmpI8vQas/run/program/evaluate_completion.py", line 150, in
if missing_pred_files: raise RuntimeError("Error: Missing prediction files! Aborting evaluation.")
RuntimeError: Error: Missing prediction files! Aborting evaluation.
I have checked my submission file beforehand using the semantic-kitti-api ‘validate_submission.py’ script and it reports that everything is okay and ready for submission (1:1 correspondence between input and prediction files).
I also noticed that the number of voxelized input files (.bin) between the old data version and the new V1.1 version has changed. Previously, sequence 21 had 523 .bin files, now it is only 432 .bin files. Is this reduction intentionally?
Thank you again for your support.
EDIT I just noticed that CodaLab allows me to looked at the error log which lists the missing files (log is attached). All of them are in sequence 21 and none of the indices seems to be actually present within the V1.1 voxel data that I just downloaded. So I guess either frames have gone missing for the official download file or the evaluation is still based on the old data.
Issue Analytics
- State:
- Created 3 years ago
- Comments:6 (6 by maintainers)

Top Related StackOverflow Question
Finally, the evaluation seems to run through and the submission gets scored. Let me know if you still experience problems.
Don’t worry about the predictions. If the original 523 frames of sequence 21 were intended to be used, it might be useful to fix the voxel data download and benchmark data altogether.