Xenova

/

vit-gpt2-image-captioning

like 25

Hello. My task is to integrate the most optimal mobile model (in terms of speed and quality) that can generate descriptive text (caption) from a photo.

I have implemented the solution from this repository, but I was not satisfied with the speed.

You have a lot of experience in this field. Could you please tell me what you would do in my place to solve this problem? Which model and format would you choose, and how would you integrate it on mobile device?

I would greatly appreciate any recommendations for solving my task!