The problem Intruder Recognition solves

In our country, we have heard and seen many cases where an unauthorized
person(broker) easily enters a room in a government office where sensitive information
of citizens are stored. The work that is to be done by a government official is rather done
by that broker in a few minutes provided that he/she is given some bribe. For example, in
government offices like Land Management and in the Transport Management office, we
highly see the above mentioned problem. So, we decided to solve this problem by
building an AI model that recognizes and stores the frame of the video when an intruder
enters the room. The CIAA officials can regularly check if any intruder has been detected
in a government office, and if detected, necessary action could be taken against the
office. This ensures reliable service delivery to a normal citizen. (P.S: This system can be
implemented in any organization that wants to prevent intrusion)

Challenges we ran into

The baseline model that we built was not giving a better accuracy. Moreover, it was
capable of detecting known faces , but failed when an unknown face appeared. Similarly,
when a new employee needs to be added to the known faces list, the model needs to be
trained from scratch which is computationally very expensive.