DeVRF icon indicating copy to clipboard operation
DeVRF copied to clipboard

test on customer dataset

Open hanxuel opened this issue 1 year ago • 1 comments

Hi Jiawei, thanks for your great work. I would like to test your method on my own collected dataset. Could you please release the code about how your get the dataset including pose, flow etc. for real world forward-facing scene?

hanxuel avatar May 18 '23 11:05 hanxuel

Hi Hanxuel, Thanks for your kind words!

We used existing released codebases for preprocessing. Specifically, we first used COLMAP (https://github.com/colmap/colmap) to calibrate the camera poses, and get the processed data files using LLFF codebase (https://github.com/Fyusion/LLFF/blob/master/imgs2poses.py). Then, we used RAFT (https://github.com/princeton-vl/RAFT) to estimate the flows, where we use the pre-trained model on Sintel dataset. Hope that clarifies!

Please let me know if you have further question.

Jia-Wei-Liu avatar May 21 '23 04:05 Jia-Wei-Liu