Thanks for your code. It seems that the threshold for pseudo labeling is 0 in utils/parser.py. Usually, the threshold is 0.95 for FixMatch-based methods. Why the threshold is so low? Thanks!
I‘d like to share my views, probably because the inner is filtered by the ova classifier, And I tried to change the threshold and found that performance degraded when the threshold is increased. Hope to help you
Thanks for your code. It seems that the threshold for pseudo labeling is 0 in utils/parser.py. Usually, the threshold is 0.95 for FixMatch-based methods. Why the threshold is so low? Thanks!