How to get Cont_fea

[yingyu13]

Hi, thank you so much for sharing your data and code! I wonder how you get the Cont_fea1 to Cont_feaN. Do you pretrain the Cont_fea layers on category-specific subset? It will be great if you can help me, thank you a lot.

[aimerykong]

Hi, thanks for your interest! You are right, we train a branch for content classification over the same dataset. Specifically, AVA dataset has content labels, while our AADB does not contain category annotation. But on AADB, we cluster the image over feature from layer 7 of AlexNet. In literature, it shows high correlation w.r.t image category on those layers from a pretrained model.

…

On Thu, May 10, 2018 at 6:38 AM, yingyu13 ***@***.***> wrote: Hi, thank you so much for sharing your data and code! I wonder how you get the Cont_fea1 to Cont_feaN. Do you pretrain the Cont_fea layers on category-specific subset? It will be great if you can help me, thank you a lot. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#17>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGKZJPkYC-uQlW7HzcCpB6pycL6B4dNkks5txELJgaJpZM4T56AG> .

[yingyu13]

Thank you for your reply