Training with dense maps

[poornimajd]

Hello great work @AlessioTonioni and team!
I am trying to use the work for training on custom data which has dense depth maps as shown below:

These are not similar to that of kitti wherein the ground truth depth is obtained from lidar point cloud and is sparse.
So can I use the depth map as it is for supervised training or do I need to make some changes?
Also if I want to use unsupervised training, I need to change the following line :

Real-time-self-adaptive-deep-stereo/Train.py

Line 99 in 33c3b92

full_reconstruction_loss = loss_factory.get_supervised_loss(args.lossType,multiScale=True,logs=False,weights=args.lossWeights,max_disp=MAX_DISP)(predictions,inputs)

with this line:

Real-time-self-adaptive-deep-stereo/Stereo_Online_Adaptation.py

Line 70 in 33c3b92

full_reconstruction_loss = loss_factory.get_reprojection_loss('mean_SSIM_l1',reduced=True)(predictions,inputs)

And to avoid other changes if I just give blank images as depth map will that work for unsupervised training?

[AlessioTonioni]

So can I use the depth map as it is for supervised training or do I need to make some changes?

You need to convert it to disparity map, but yes you can provide dense data. Invalid points (for example the black area in the uploaded image) should be flagged with a 0 disparity and the code should ignore them.

Also if I want to use unsupervised trainin

Yes if you want to do unsupervised training from scratch.