ARML

This project demonstrates how to identify faces using Apple's new Vision,CoreML and ARKit APIs.The code is written in Objective-C.

Requirements

Camera Hacking : Since ARkit uses a fixed-lense camera to render the screen, the camera won't be able to auto-focus by itself. To tune the camere you need to get access to the AVCaptureDeviceorAVCaptureSession . However, this is not possible in ARKit as described here . I solved this problem by accessing the availableSenors property of ARSesion in runtime and find the ARImageSensor object which holds the reference to the AVCaptureDevice and AVCaptureSession instance.
Machine Learning: To identity different people we need a pre-trained CoreML model. You can use caffe or other neural network infrastructure to train your model. For this demo, I use the Mircorsoft's Custom Vision Serivce which is free and convenient to train images online and you can also download the reslut in CoreML model format.

Link to CoreML documentation
Link to Apple WWDC videos, samples, and materials for information on CoreML and Vision Framework
Link to Custom Vision Service Documentation