What is pretrained-model?

[buimanhlinh96]

Hi @wang21jun ,
Thanks for your contributor!
In your submit at https://github.com/JDAI-CV/FaceX-Zoo/tree/main/training_mode, i see you had two kind models: SOTA backbones model and SOTA heads model. What is the difference between them?
Thank you!!

[wang21jun]

SOTA backbones model : models which are trained by the SOTA backbones with the same head, i.e. MV-Softmax.
SOTA heads model : models which are trained by the SOTA heads with the same backbone, i.e. MobileFaceNet.

[buimanhlinh96]

@wang21jun I got it!
Thank you for your clear answer.

[buimanhlinh96]

If I just want to train head, not re-train the backbone in my custom dataset. How can i solve it?

[vietth-bka]

If I just want to train head, not re-train the backbone in my custom dataset. How can i solve it?

You can search keyword freezing backbone pytorch.
Specifically, you can set requires_grad = False for the whole learnable parameters of the backbone. But the easiest way is just putting the parameters of head into the optimizer but not the params of backbone.

[buimanhlinh96]

@vietth-bka thank you!