Hello, it seems that there are multiple files in the test data directory with no acceleration inputs.
For example, public_data/test/00015/acceleration.csv has no rows, even though in the sample submission files it is listed as being 15 seconds long. Is that expected behavior?
Hi marcvshdr,
Yep. This is expected behaviour.
One of the interesting challenges that this competition is trying to tackle is that of missing data in real sensing/IoT systems. One of the challenges that will be faced is missing data from sensor streams.
In the blog post that I prepared I showed a simple data imputation method
http://blog.drivendata.org/2016/06/06/sphere-benchmark/#Preliminaries:-data-imputation
wherein I used the
sklearn.preprocessing.Imputer
class to fill in the missing data.
Thanks for your interest! I’m happy to answer any more questions that you might have.
Niall
Oh, that’s OK, no worries.
The input is still a bit misleading (for some of the test videos, for example, I’ve discovered that there are no readings at all). There should be a something like a line called:
As a comment right at the beginning of the file or such. As feedback for next time.
Thank you for your suggestion.
Best of luck in the rest of the competition! 