Hello Sir, my answer is 5. haha
Let's discuss about your project.
As for motion detection, there are a lot of OpenCV default method for that like backgroud substraction method.
But I don't like these method because they are almost low in accuracy. so I recommend other motion detection and tracking method by using Deep learning Neural networks.
Fast-RCNN and Yolo are my favorite. but in order to use these methods, we need to use tensorflow or keras-based environment on your device. I think there is no problem to realize these ideas so far.
As for identifying user id, we can use various object detection and tracking algorithm. in order to train the detection model, I need to have at least 10 images per person.
At last, in order to send the id to server, it is necessary for me to create daemon server where all detected ids would automatically be uploaded.
If you are willing to hire me or have any questions, please ping me.
Best Regards
Andrey
PS: as for timeline and budget, I can finish off this project within $500 and 5 days.