Background
We would like to create a contact book which could be add real-time. It is basically the ultimate contact book. The idea started when we would like to try to detect face and it attribute to generate a model for each user. It fell upon a unique system for adding face data and his/her attribute. So, it would be able to sync up with our face database and the person face.
How It Works
Scanning
- In the interface, we’re using Node JS to send data to Python Flask
- Python Flask then compare the picture image it receive with our face database
- We create a embedding of the image to be compare with our SVC (Support Vector Classifier – a subset of SVM ~ Support Vector Machine) which has been fitted with a linear kernel.
- It would output the person data which is save in TXT file
- Falsk would send back the result to Node JS
Training
- In Training mode, Node JS would send 20 pictures to FLASK
- It would be split 80% Training & 20% Testing
- Then, It’ll use MTCNN to brake down the images to their face
- Using Facenet to get the facial features
- After that, create an embedding of the faces
- Using SVC to train the model
- Lastly, Add all of it to a single save state file
Problem
- We face an error in call thread function when combining flask and tensorflow
- We solved it by calling the function via os.system(“”)
Future Works
We would like to connect it with social media API. Then, we could compare their facial data with the profile picture. Afterwards, we ask like is this the person your looking for.
Pictures
Team Members
Yoksan
Longlong
Ivan
Link
Code :
https://drive.google.com/file/d/1BzYfJZPkoyOnLcVSunSVA56CZBHG1Sds/view?usp=sharing
Presentation :
https://drive.google.com/file/d/1436KBpAeyKhypgvkXGDJwDRmfTnJ15kR/view?usp=sharing
Video :
https://drive.google.com/file/d/1436KBpAeyKhypgvkXGDJwDRmfTnJ15kR/view?usp=sharing
