YOLO ALGORITHM- uniqueness of objects

edit

and how is one thing related to the other ?

what would your "discriminator" be ? (apart from probabilities it's already returning ?)

you get bounding boxes, so you have the "geometric center". for the center of mass, you would need contours or binary crops

(maybe you need a more sophisticated cnn, like RCNN, where you could do something with segmented object masks ?)

"uniqueness of objects" -- please explain !

@berak have put two web video cameras on the corners of the ceiling of room. Lets say that two people enters the room.I grab the first frame of the first camera and the first frame of the second camera. Afterwards, i implement YOLO algorithm on these two frames separately. So i have: Coords of geometric center1 , center2 (first camera - person1 person2) Coordsof center1',center2' (second camera - person1 person2) I intend to use cv2.triangulatePoints() in order to find 3D coords using coords of the centers. In order to do that i must know the corresponding geometric centers. That is, i must know that in the first frame of the first camera there is person1 and person2 and not just person and person( same thing for the second camera) so i can use the above function and find 3D coords

so really, you're unable to find the center of a given bounding box ? (maths fail)

@berak my question is:

how can i have the coordinates from the geometric center of one object on the 1st frame on the 1st camera and the geometric center of the SAME object on the 1st frame of the 2nd camera through the results of YOLO algorithm

SAME objects (from 2 seperate detections in 2 images), that's the key now

@berak exactly, i pointed out in the 1st comment, saying corresponding. Anyway is there any solution to this problem?

no i don't think, there's any solution, your original problem seems to be ill-stated.

(you'll never get a proper 3d pose from the input you have now)