Samsung J5(SM-J500M) was used to test the app as a low powered board was specified in the requirements
The app seems to be very innacurate, currently working on debugging at the moment. I don't think it is the model, it could possibly be the low quality bitmap it reads in for inference as compared to the high quality x-rays it had to train on.
https://blog.tensorflow.org/2018/03/using-tensorflow-lite-on-android.html