Facial Recognition for Home Security

Objective

To build a functional Face Recognition system that can open a door, that was for the beginning, now the new objective is to make a facial recognition system and make work through a raspberry pi, so we can implement it in as many things as we can.

Meeting

coming soon

Students

Mentor(s)

Blog

Week 6

August, 05 2019


●Try to make The python code to work on Window but haven’t had any good results. Try to use Anaconda, Minianaconda for python and download from console and from interface yet the code outputs That some parameters need to be filled yet it dose;t let us input while in mac it is allowed.

●Work with the raspberry pi and the breadboard to make sure that the circuits work properly by doing a little led light experiment and it works perfectly.

Week 5

July, 29 2019


●Try more samples on the Facial recognition system to see if we can increase the accuracy by increasing the sample to 50 and there were no improvements.

●After some research we found out about the method we were using called Hog which works with a small data set, is faster but less accurate.

●Found out about CNN which is more accurate, yet it is way slower and in order to work need a huge dataset. Now we have to find better ways of implementing it taking into consideration the limited processing power we have.

Week 4

August, 07 2019


● Do to the fact that we haven’t been able to connect properly to azure and the fact that every transaction was charging us money, we are now We decided to use OpenCV but trying to decide which programming language we should use for it, C++ or python.

○ C++ Advantages: Both of us Have more experience with it and it is easier to work with in Visual basics, but it is harder not adaptable.

○ Python Advantages: the code can be used in Apple and Windows devices without a lot of changes and, because the two members have a different operating systems, this will simplify the testing between both computers.

● Successfully downloaded the libraries on python and used a sample code to make facial recognition work using this source code, Trained it with samples of 3 different people putting 21 pictures each, it works, and is detecting pictures now.

●Also successfully made live video detection but is detecting an unknown person as someone it knows which shouldn’t happen.

●Another issue is that the code is not working on windows, just on mac.

Week 3

July, 15 2019


● Azure API Still not working properly, The key that we receive to have access to the cloud is not being valid, and the use of OpenCV to do the FAcial recognition offline, even though cheap, may bring performance issues when used in the raspberry pi, our goal for this summer has changed to being able to detect faces in live video. Even though we are still aimg to open the door we figure out that we can do more than that so we are aiming to just be able to detect faces and after that expand on as many applications as we can.

Week 2

July, 08 2019


● Installed Microsoft IOT on the raspberry pi so that we can connect to a remote computer to the raspberry pi so that it does the process, and then tells the raspberry what to do.

● Started looking for other methods for the use of facial recognition just in case that Microsoft azure does not work as well as we planned. In the process found out about Opencv an open source API used for image recognition, it can be used in may softwares andis free in contrast to Azure but because it not cloud based, we will need a way of processing the information in a decento computer and sent the results to the raspberry pi.

Week 1

July, 01 2019


● Started researching ways to do the project resources for the project.

● Our overall plan is to connect the camera and Microsoft azure which is a cloud API service with cognitive services capabilities that allow us to work with face detection, voice detection, etc. using this we confirm the face and tell the raspberry to open an electrical door lock if it is a valid person.