face_recognition for Android

[mmelatti]

• face_recognition version: latest
• Python version: 3
• Operating System: mac OSX, Android, Linux

Description

I'd like to add face recognition to an android application. I'm particularly interested in the face_encoding 128 floating points that this api provides. I have Android apps running OpenCV. I'm trying to decide the best approach and if anyone has experience with this?

• Can I take the "Training Data" from this api used to generate 128 floating points encoding to get same face_encodings in Android natively, (I did something like this with OpenALPR License Plate Classifiers .xml)
• or has someone else had better experience with compiling python code as an .apk (with all of the requirements included in this api) (grabbing camera, sensors, internet from android device)

I would want to send face_encodings to a backend server (spatial database) to have a more distributed face recognition platform.

Thanks!

[vearutop]

https://github.com/tzutalin/dlib-android-app this lib might be similar. https://github.com/gv22ga/dlib-face-recognition-android this looks relevant too.

[rw86347]

Did you ever get anywhere with this?

[mmelatti]

@rw86347 Yes, check out my public repo:

https://github.com/mmelatti/face_recognition

I'm using postgresql configured for 128 floating points. Good for face databases containing up to around 80 million faces. (More testing required)

I have examples and use cases included.

For android I do basic face captures then TCP images to backend for processing and query face database, and send results back to device. Some image processing could take place on android device instead of server but this logic currently resides in python and uses python libraries.