Start the camera, make sure every student's face is clearly visible, then scan the group.
The server detects all faces in the captured frame, checks three camera frames, requires a consistent match, shows a 3 second NO option for each match, and then marks each accepted student once. Unrecognized faces are reported separately.