Number of videos and number of rows in train_labels.csv and submission_format.csv

Why is there 456002 files in nano and micro sets, but only 204130 rows in train_labels.csv?
How many rows should I submit?

Also I noticed 87484 rows in submission_format.csv file. What does this mean?


You’re seeing everything correctly. For technical reasons, we provided more video downloads than are actually used in the competition. The total videos in training and test are 204130 + 87484 = 291614. The filename columns in the training and test csv files will allow you to grab the ones you need for both training and inference. Feel free to use the rest, but no labels are provided!

See our benchmark for more info!


1 Like