Closed zhilaAI closed 3 years ago
Hi, You can both: 1) save depthstilled flow maps as flo file, instead of png 2) change the way you load the ground truths in your data loaders, as in RAFT
Thanks for your response. For the first option, I tried but OpenCV and Matplotlib do not support the .flo format. I do try the second option.
Hi,
to read .flo files you can use the readFlow
function defined here:
https://github.com/princeton-vl/RAFT/blob/master/core/utils/frame_utils.py
Hi, Ya, I checked the code and saw the readFlow function but the problem is writing files in .flo format. Thanks for your suggestion.
Sorry, my misunderstanding. From the same file, have you tried the writeFlow
function?
No no thanks for your help. Yes, I tried to use writeFlow to convert png to .flo format but failed.
Did you try it to save flow_01
or flow_16bit
?
I tried flow_16bit
Ok, that one is already converted to the KITTI format. You should try with flow_01
I'll try the flow_01 file.
Thank you.
Hello, Thank you for your great code. How did you train the models for optical flow where the ground truth should be in .flow but your code generates .png format?
Thank you.