All geek questions in one place

What makes object representation and recognition tough?

Intuitively, it would seem that given the dozen two-dimensional images from different angles of almost any object, it would be easy to build a three-dimensional representation of this object. Subsequently, the library of three-dimensional representations obtained in this way can be used to identify new 2d images.

What literature exists in this direction and why has it not yet caused strong recognition of objects?

+3
source share
3 answers

What you want to do is really possible, but (there are quite a lot of them)

for 3D reconstruction:

  • For anything but the simplest forms, you need more than just a few dozen images.
  • , , , , .
  • , ( ).
  • (.. ) 3D- , (, ) , .

, , , , 3D-.

, , 2D- , , . , , 3D-, , .

, 3D-, .

Photosynth . , , .

, , ! , , , , - , , ( ) . .

2D- SIFT. , 3D.

+4

"", . , , , . ( 10 , 10 ^ 14 , , ). , , , , . .

, , - , "". , , "" . , ( , , , "", "", "" ) , , , . , " " , , .

+7

, , , . . , Google Goggles , .

. 2D- .. . , , , , . .

, , 100 . . , .

, 5- . , . 100 3 1 = 300 /. 30 , 31 556 926 5 , 1,4 (1.4x10 ^ 18).

+2

Source: https://habr.com/ru/post/1793960/

More articles:


All Articles