huawei-noah / benchmark Goto Github PK

Does the attention_dim feature in the implemented AutoInt model correspond to the number of hidden units d' in their paper? In section 5.1.4 of their paper they have set d' to 32. So I am wondering if this corresponds to attention_dim = 8 in your code.

Dear Authors,

It seems some SDRTV-HDRTV pairs are missing. As it is mentioned in the paper, and there are 97 4K SDRTV-HDRTV pairs in the test set.

Can you share with me the whole test set? If you are considering releasing the code of this paper, please let me konw.

Best Regards,

请问可以提供数据集下载地址吗？感谢

Dear Authors,

It seems the code for the MB-STR is missing in your repository, although it is mentioned in the paper itself, and there was a URL pointing to this repository.

Can you share with me the source code and the data of the proposed model?

Best Regards,
Ahmed

In the inception block of HA-Conv, the channel of the feature map is halved along each track and then restored to the origin by concatenation. As descripted in the paper, the output channel number of the third HA-Conv is 3 which cannot be divided exactly by 2. How do you resolve this conflict?