Alessio Tonioni

> > > `X` is not supported. Use `x` instead. Besides 'X' and 'x', with any phase that is not a multiple of 0.5 the signal does not start from...

How have you serialized the predicitons to numpy? Have you rescaled the input images?

As far as I remember there was no scaling in the tensorflow version of the code. If you are only interested in inference with dispnet you can also refer to...

Are you sure to be using our implementation of L1Loss layer? One of the difference between our code and the original flownet one is indeed the number of bottom blob...

If you are using our train.prototxt the model is trying to minimize only flow_loss1 (i.e. the higher resolution disparity) so I think that your model is working. If you take...

Hi, Both the normal supervised version and the ICCV adaptation are implemented in the tensorflow version, however, to use the adaptation loss you will need to provide disparity map and...

In the unsupervised version "path_ground_truth" is the path to the disparity map obtained with classic stereo algorithms (SGM or CENSUS on the paper) over which you compute the confidence measure....

Hi, Regarding 1 what do you mean exactly? The details on the dataset preparation are reported here: https://github.com/CVLAB-Unibo/Learning2AdaptForStereo#data-preparation. It boils down to having one CSV per stereo sequence with consecutive...

HI Tiam, sorry for the late reply. I honestly cannot remember exactly how the models were trained or performing, but a 99% D1 error seems quite wrong. Do you get...

Hi Tiam, I honestly don't remember anymore the range of disparities used and if it's fitting with the one you measured. In general the range should more or less match...