What is the first dimension in log_prob in learn.TTA()

Hi, during the prediction of the Humpback Whale Identification Challenge challanged, when I finally run log_pred,y = learn.TTA() and print the log_pred shape


It’s is giving me the following as output

(5, 1808, 4250)

Now, I understand that 1808 is the number of test images and 4250 is my number of classes, what I don’t understand the first dimension i.e. 5. What is that this first dimension signify?

The 5 dimensions are the 5 augmentations per single input for the predictions.

It is a uneven number to always have a yes/no result (and not a 50%/50% result).

Best regards

Thanks, @MicPie. :grinning: