Datasets in training and schedule

[orram]

CREStereo/train.py

Line 147 in ad3a161

dataset = CREStereoDataset(args.training_data_path)

Thank you for supplying this code and training procedure!
In the paper (and the git readme), you say you train using other datasets as well ([SceneFlow], [Sintel], [Middlebury], [ETH3D], [KITTI 2012/2015], [Falling Things], [InStereo2K], [HR-VS]).
Yet, in the train.py, you only refer to your CRES dataset.
Can you elaborate? Are you training on other datasets before? After?

Thank you!

[ynma-hanvo]

yeah, the public dataset is really too hard to use. can share the tools you used to generate the cres dataset from public dataset?