Have developed Facial Recognitions System for our client in India, I am comfortable with:
backends = ['opencv', 'ssd', 'dlib', 'mtcnn'] FACE DETECTION
models = ["VGG-Face", "Facenet", "Facenet512", "OpenFace", "DeepFace", "DeepID", "ArcFace", "Dlib"] FACE RECOGNITION
And i think if your project involves doing something original then it will be fun to work with you.
Few Challenges : Usually most systems need you to train entire model again if you just add one more image to the database, which is kind of re-inventing the wheel and no one likes to have such a pathetic pipeline.
2- Frequent guests etc who visit the premises are again required to submit 3 pics which again is not kind of good impression to any one, after all they are not criminals, right.
Frankly speaking most coders use code from Github without thinking that the code is meant as a guide not the entire solution for the real life challenges, hope it makes sense, Let me know , if this makes sense , and we can understand things better.