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

inconsistence about final feature between paper and code? #2

Closed hiyijian closed 5 years ago

hiyijian commented 5 years ago

Dear @iamhankai , Final feature equals region feature plus attr feature in your paper, while its equals concat of the two in your code. Is there any experiment shows which is better?

iamhankai commented 5 years ago

Dear @iamhankai , Final feature equals region feature plus attr feature in your paper, while its equals concat of the two in your code. Is there any experiment shows which is better?

Yes, we have tried two manners, and there is no significant difference.

hiyijian commented 5 years ago

@iamhankai Thanks