The Google Cloud Vision API offers following features- Landmark Detection, Logo Detection and Optical Character Recognition, in addition to the Label Detection feature.
These can be useful to fill the alt text/title text by default.
These features would be used to fill the alt text of the image file based on the choice given by the end user.
The user can check the checkboxes present in the Edit form of the Image File entity, and on its basis, the alt text would be appropriately filled.
Create a patch which would fill the alt text of the image file entity.
(This is one of the tasks of my GSOC 2016 Project.)