Combine both image and text features

[ziyanyang]

Hi,

I have checked the tutorial for NMT, and I'm wondering if there is a way to combine both text and image features to train machine translation models? For example, https://www.aclweb.org/anthology/P17-1175.pdf combines image feature and text feature to decoder. Does current OpenNMT support such operation?

Thank you!

[francoishernandez]

Did not check in details but you might want to have a look at #1597.
Also I gave a few pointers to a similar question here.