Down sizing the predicted depth

[maherkarim]

Dear,
Thank you for your efforts.

I was testing your code using my own data. I have prepared the data and it was loaded correctly.
When doing training, I am receiving always this error message
The size of tensor a (196608) must match the size of tensor b (12288) at non-singleton dimension 2

My images, and depth are scaled to 384*512
The network output for the fake_z is torch.Size([4, 64, 96, 128])

and the model is not able to train. I even made the depth images bigger 4 times to consider the downscaling but got the same error.
can you explain how to do training with the images specified above. or direct me what am I missing

Thanks

[tomashlehmann]

Did you solved that problem?
I've got the same error...