Training with negative and positive disparities

[Andyjk1406]

This network is excellent!
I'd like to train on my own data, but my rectified images have positive and negative disparities.

I see you use a max_disp, but is it possible to use min and max disp (and create the volume sizes based on the difference between max_disp and min_disp?

I suppose the alternative is to shift all of the right images to the right by abs(min_disp) but this seems less elegant and will enlarge the images

[Andyjk1406]

Just to add to this, other training data may have a min_disparity that is much larger than 0 (eg disparity range from 100 to 250). Again I presume this would mean a min_disparity option would be more efficient than sliding the whole right image to the left