Hello,

It sounds nice in theory. But knowing the current limitations of object recognition and our category/tagging system on Commons, I don't see how it could really help to improve the project. For example: We don't tag every object in the background since it is much better for the user to find images inside the categories that have this object as the main motive. In case of something like a ball it would need to correctly distinguish between a football, a globe, marble, melon, etc. If it guesses wrong and would add the tag it would accidentally increase the efforts for our maintainers.

Maybe you want to elaborate in more detail what possible use cases you have in mind, where it would be a real help for our contributers. I like the idea, but at the same time I see a lot of difficulties to transform it in a realistic improvement for the project.