search

iamhankai / attribute-aware-attention

[ACM MM 2018] Attribute-Aware Attention Model for Fine-grained Representation Learning
https://arxiv.org/abs/1901.00392
156 stars 30 forks source link

how to train without attribute label #4

Closed twmht closed 5 years ago

twmht commented 5 years ago

Hi,

from the table(1) in the paper, you mentioned that the network can be trained without attribute label.

in such manner, what is the form of the final loss function? Do we only need to care about the categorization loss?

iamhankai commented 5 years ago

just remove the loss of attribute classification, or set lambda to 0 in https://github.com/iamhankai/attribute-aware-attention/blob/ab3210705524bcb56f0121de0bcd47f3f93fb6c5/cub_demo.py#L98

twmht commented 5 years ago

very interesting. I am going to do the experiment. How about the other lambda? Do they remain unchanged?

iamhankai commented 5 years ago

very interesting. I am going to do the experiment. How about the other lambda? Do they remain unchanged?

Yes, unchanged.

twmht commented 5 years ago

close.