Closed hengfei-wang closed 3 years ago
I have the same problem with the green screen, if you solved the problem, please reply)
I did something wrong in the transformation of camera poses. Have corrected it.
closed.
@hengfei-wang Hi, i have the same situation. Could you tell me how to use calibration parameters of cameras to create poses_bound.npy? And would you like to share your code? Thank you so much!
@Duncan1115 Have you figured out a good way to solve it?
can any one help me in training part of nerf existing code?
Hi, thank you for sharing your impressive work.
Recently, I am trying to train nerf using my own data. Since colmap didn't work on the pictures I collected, I created my own poses_bounds.npy using calibration parameters of cameras according to this the following step which is from link:
I also see the issue about the unit of data in poses_bounds.npy and set hwf in pixel and set translation and bounds in meter. But I get some confusing result like the following.
https://user-images.githubusercontent.com/39575336/112758030-c4cb2b80-901e-11eb-89d4-26f66a480471.mp4
The training picture is one person sitting in front of a green wall. It looks like that the training result mixes some colors from the person and the wall but makes no sense. The person is sitting at approximately one meter distance in front of the cameras so I set near and far bounds to 1 meter and 2 meters. Is there anything wrong I have done? Hope someone can kindly provide some help.