When i am testing model with my own stereo dataset it was not giving depthmap correctly?even after calibration/rectification?

could anyone please explain does this deep learning models will work for my own stereo cameras?does these deeplearning models give depth map correctly or else it only for online available datasets and if i want to train my own dataset-what is the procedure?how to check does the depth map correctly giving depth in real world meters?