-
Notifications
You must be signed in to change notification settings - Fork 73
Pose convention
Pakkapon Phongthawee edited this page May 28, 2021
·
1 revision
NeX is using OpenCV convention while NeRF and LLFF are using OpenGL convention.
( Graphic taken from What are the coordinates? )
The orange point show below is the same point. However, the different conventions make the value different.
For OpenCV to OpenGL, we have to Flip the axis.
The orange point in OpenCV [x,y,z] become [x,-y,-z] in OpenGL convention.
We can multiply this 3x3 matrix to convert the coordinate of the point between OpenCV and OpenGL convention
Let's R be a 3x3 rotation matrix such that
We will do these following step:
- Convert the point [x,y,z] into OpenGL coordinate
- Rotate (Apply) in OpenGL coordinate with the rotation matrix
- Convert the point into OpenCV coordinate
Let's do step by step multiplication.
Here you get the line 324-325
For line 323, you just flip the y-axis and z-axis of the camera center.