Predicting random images other than KITTI & NYU, when the model is trained on either of the dataset

[Oreobun]

Hi all I am pretty new to python and the whole field of monocular depth estimation, I have successfully run the bts_test.py on the NYU dataset and would like to ask if the model can be used on predicting other random images outside of the training/testing data. Some questions I have are how to achieve it? Am I doing something stupid and this is not how monocular depth estimation works? Or it is possible to run on random images but with a much lower accuracy? Please enlighten me, thank you!!