-
Collect data of every employee (5 Images per employee with slight variations) according to the constraints mentioned in "Dataset Collection Parameter.pdf" document in the root of the repository.
-
Put the collected images (Not the meta files; they are only for reference) in the ./data directory and check whether the recognizer works well on it. If the performance of the recognizer degrades, try different approaches for recognition.
-
For the detection part, HOG or Haar Detectors are sufficient for the project. If a solution is required for the more open-ended problem statement, check the ./misc directory for details of models already tried for detection.