Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

It's easy to trick by just having your hand anywhere in front of your face, even a foot away. Is this something current image classifier architecture could address with more data? It seems like it'd be hard to tell whether a hand is large and on a face, or small and far away from the face unless there is some sort of depth estimation going on (based on cues like shadows).


Is this a problem? Do you find yourself waving at your webcam much?


I find that my hand goes between my webcam and my face reasonably often.

It's still a brilliant idea and I love it, regardless of its limitations.


Maybe, but its actually really good at not firing when you scratch your head, which is probably more common than casually waving at your webcam ;)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: