How can i get 3D world coordinates from 2D points

Hi, i just started in the computer vision world, and now iam researching about keypoints.

I just extracted keypoints and i cant get their 2d points in the image, that is easy, but also i want to know if its possible to get their 3d coordinates (x, y, z) in the world, so then i can use solvePnP and get the camera pose with keypoints.

I have calculated camera intrinsics, but i dont know the extrinsics.

sounds, like you already got something wrong:

to use solvePnp() you need an existing 3d model (no, that’s NOT the result of it !)
IF you have coresponding 2d points, you can get a 3d pose from that
(think: a 3d face, and 2d face landmarks for the 2d points)

those are arbitrary in nature – it’ll lead nowhere (no corrspondance)

Can you explain this in more detail? Think that im just starting hahaha.

So then, you mean that its imposible to determine 3d world coordinates for keypoints and for the camera in this case?

sorry for the sarcasm, but it’s all in here

(read a book, don’t ask)

Nice help. I hope that when you have a problem people will give similar answers to you.

Then, do you mean that its imposible to get the localization of the camera using keypoints?

“keypoints” is a broad term.

“facial landmarks” are a more specific term… if that is what you have.

you haven’t explained what your situation is.

be very specific in how you present the problem. use pictures rather than words.

and yes, you seem to ask questions that are answered by existing literature (even opencv documentation…). you need to realize that you are learning. your attitude caused that reaction.

I dont have facial landmarks for now. Just taking pictures with my webcam and testing python classes to extract keypoints, like Sift to get all of those keypoints and compute descriptors.

Just want to know if there is some way to get the camera pose, 3d coordinates or whatever to know where those keypoints are in the real world. Sorry for not explaining so well but, as you said, im learning and there are so much things that i dont understand, thats why i came to this forum, and i think that when you started maybe you where as confused as me, even reading docs and literature.

And yes i was reading about homography and those things, but i just need an explanation for “dummies”

browse the tutorials in the OpenCV documentation.

you’ll find a whole section on 3D stuff. you might find the parts involving ARUCO interesting.