Where is the code of Auxiliary visual-semantic matching?

[huynhtruc0309]

Hello. I cannot find the code of the Lvs loss. Please help me. Thank you.

[yanbeic]

Hi,
This part can be done by pre-training the base network using the same loss, but the different texts (i.e. tagged captions in fasion200k). So the difference is only to replace the text input.