I was wondering if adding a lightweight machine vision model to images would be helpful. I was trying to search up "jaguar -car -football" (without the quotes). And many many cars still appeared. Would it be feasible to use the machine vision model to still remove the cars? This would be to improve search image results.
So when you use a -"insert object/person/thing" it can remove the pages that have that thing. But it only kicks in when i want a particular thing removed. This would be so it isnt constantly scanning. I understand if this would be a fairly intensive thing and cant be done but i felt like it warranted mentioning.
