When bounding-box is on person A/B/C click respective "register person" button multiple
times (10-20) with different poses of face to add more instances (gives high accuracy)
Note: Currently NOT using powerful backend in this prototype. Instead simple naive model is being used.