Good practice

More images need to be downloaded, both of people wearing masks and helmets.
Images of people wearing helmets without exposing their faces preferred, i.e. their face should not be seen.
After sufficient images have been downloaded, both TF and TFLite models should be retrained. "TFLite retraining" meaning that the retrained TF model should be converted to TFLite model using existing code in repository

This can be faster, but may require more code and add extra steps. TBD.

This screen should include the two or three questions present in the code, with answer selection done by radio buttons (yes/no).
Answers can be saved in a text file and uploaded on the server, which will be monitored by the system in Pi.
When Pi receives the answers, code conditionals will be executed appropriately. Meanwhile, the app returns to the main screen. (Cannot return to main screen if all questions are not answered)