what does R and T in base_cma.json mean

[buptseelihang]

if R means rotaion matrix ,why it have four parameter? what does slice and time mean

the default value of T is [0.5,0.5,2.0]? i try to change it in nerf_loder.cu but it doesn't work

how to use the transform_matrix of frames to get cameras' position and view direction so i can render a high quality video?

THANK YOU~

[Tom94]

Hi there, if you have a stored camera path, you can use

testbed.load_camera_path(args.camera_path)
testbed.render(start_t=time)

where time is a value from 0 to 1 that indicates the position along the path.

As for R, T, slice and time:

• R is a quaternion describing the orientation (R for rotation) of the camera,
• T is the position (T for translation) of the camera,
• slice is used to animate the slice plane when rendering in slice mode, or, alternatively, to animate the focal plane when rendering with depth of field, and
• time is the current time along the camera path at the time of saving it.

[buptseelihang]

Thank you for you replay ,i have got R use quaternion to describe rotation. I convert quaternion into rotation matrix but i find it's different from with transform_matrix in transforms.json.Is there a coordinate system conversion? it seems exchange R's row and change the sign in 2 and 3 col. But i can't find T's contact between camera path and transforms.json .

In a word ,how to convert transfrom matrix to camera path.

[kajc10]

@buptseelihang did you manage to solve this? I'm trying to convert transforms.json coordinates to base_cam.json to render explicitly the same frames.

[0xiNach]

@buptseelihang Answered in #490