When training a classifier, don't freeze the weights of the decoder and codebook?

Gengzigang / PCT

This is an official implementation of our CVPR 2023 paper "Human Pose as Compositional Tokens" (https://arxiv.org/pdf/2303.11638.pdf)

MIT License

311 stars 20 forks source link

Closed epleone closed 7 months ago

epleone commented 7 months ago

hi, When training a classifier, don't freeze the weights of the decoder and codebook? I didn't find any steps in the code for freezing weights.

Kun-Ming commented 2 months ago

Hi epleone,

Do you find which code indicates the weight of the decoder and codebook has frozen?

epleone commented 2 months ago

@Kun-Ming

Yes, I found it before. But it's been a long time.

It might be this line.

Kun-Ming commented 2 months ago

@epleone Thanks! That makes sense.

epleone commented 2 months ago

Glad to help you.

Tommy-Hsu commented 2 weeks ago

hi, @epleone ,

Does this method (not freezing decoder and codebook )get better accuracy ?

Cause I couldn't reach accuracy than heatmap base q_q