kuhnkeF / ABAW2020TNT

Submission to the Affective Behavior Analysis in-the-wild (ABAW) 2020 competition.
MIT License
35 stars 9 forks source link

Question about the training #7

Closed dcafarelli closed 3 years ago

dcafarelli commented 3 years ago

Hi, I appreciate your work.

I saw that you reshaped the first layer of your model adding an extra channel input and copying the weights from the "old" layer. Then, what about the weight in the last channel? Did you retrained the whole network or just the first and the last layer?

Thank you very much

kuhnkeF commented 3 years ago

Hi,

we fine-tune (train) the whole network during training. Hope this answers your question.

dcafarelli commented 3 years ago

Yes, thank you!