How to adjust the percentage of the training data with labels for NTU RGB+60？

fmthoker / skeleton-contrast

43 stars 4 forks source link

How to adjust the percentage of the training data with labels for NTU RGB+60？ #1

Closed ch1998 closed 2 years ago

ch1998 commented 3 years ago

Your code is wonderful。 In the paper，you report results whenusing 1%, 5%, 10% and 20% of the training data with labels for NTU RGB+60。 I hope to how adjust the percentage of the training data?

fmthoker commented 3 years ago

@ch1998 you can randomly select any 5% of training examples during the fine-tuning step. However for reproducibility seed the random generator.

ch1998 commented 2 years ago

I get it. Thank you!