search

thuml / OpenDG-DAML

Code release for Open Domain Generalization with Domain-Augmented Meta-Learning (CVPR2021)
33 stars 6 forks source link

Question about how to implement DAML-S #5

Closed Noguchi-Masashi closed 1 year ago

Noguchi-Masashi commented 1 year ago

Hi, Thanks for your excellent work and code!

In Appendix, You evaluate DAML-S where the three domains each have a specific classifier but share the whole backbone. I would like to know how to implement DAML-S.

Would the following changes make DAML a DAML-S?

backbone1 = resnet18_fast()
#backbone2 = resnet18_fast()
#backbone3 = resnet18_fast()

classifier = ClassifierFast(backbone1, backbone1, backbone1, num_classes).cuda()

Thank you in advance.

jis3613 commented 1 year ago

You have to alter the daml_utils.py file. You have to define a new class that only uses one backbone and multi heads.

Noguchi-Masashi commented 1 year ago

Thank you for your kind response.