Performance on other tasks

[armenjeddi]

This is a great paper! In the results I mostly see the benchmarking on OCR related tasks - how does it work on other benchmarks like general VQA, localization, counting, etc?

[mvish7]

Hi
I’ll post evaluations of a model finetuned for image captioning in couple of days.

[mvish7]

Hi
I have updated the README with results. Please note that this is not official implementation of the paper and this implementation differs from the original method in terms of image-text inputs packing.