Ask Your Question
0

Haar Cascade or ...? OpenCV detection

asked 2020-09-01 09:04:02 -0500

So i had an idea of making a haar cascade that can detect when an someone's x-ray shows something that shouldnt be there , like if he is healthy or not . I got a bunch of xrays of people with health issues and a bunch of healhy people . The thing is that as a starting thought i was thinking that it may work out well . But im kinda stuck in the thought process of it and i kinda don't know what to do. Most of the tutorials out there are about face , car , watches detections which is kinda "clear" on what you gotta do , just put some pictures of watches and some other random ones that don't contain watches. In my case , they will be pretty similar and i don't know if it will work or not and if not , does anyone have any idea on how to make it work other than haar cascades ?

edit retag flag offensive close merge delete

Comments

Everyone will answer "deep learning", Haar cascades will almost certainly not work for anything close to what you say you want to do. You need to narrow you problem, look for one specific problem detectable in X-rays. Start there and add complexity as you succeed.

Der Luftmensch gravatar imageDer Luftmensch ( 2020-09-01 09:28:23 -0500 )edit

Probably a unique test for each of broken bones, infection etc. so as not to pollute sample pool. Then proceed with overall health computation.

There's clearly no quality standard - just look at the app store or search for articles on ethics. Unfortunately, if yours turns out awful you could re-brand it as a silly learning toy for kids.

kpachinger gravatar imagekpachinger ( 2020-09-01 09:30:14 -0500 )edit

1 answer

Sort by ยป oldest newest most voted
1

answered 2020-09-01 09:38:52 -0500

berak gravatar image

that can detect when an someone's x-ray shows something that shouldnt be there

clearly, no case for haar cascades.

try to find out, what "binary classifier" means

(and please, do your own research before asking)

edit flag offensive delete link more
Login/Signup to Answer

Question Tools

1 follower

Stats

Asked: 2020-09-01 07:45:49 -0500

Seen: 43 times

Last updated: Sep 01