novatecconsulting/facerecognition-in-arkit

Detects faces using the Vision-API and runs the extracted face through a CoreML-model to identiy the specific persons.

FaceRecognition in ARKit

This is a simple showcase project, that detects faces using the Vision-API and runs the extracted face through a CoreML-model to identiy the specific persons.

Requirements

• Xcode 9
• iPhone 6s or newer
• Machine-Learning model

Machine-Learning model

To create your own machine-learning model, you can read our blog post "How we created our Face-Recognition model"

The short version is:

• We trained a model in the AWS using Nvidia DIGITS
• Took a couple of hundred pictures of each person, and extracted the faces
• Also added an "unknown" category with differnent faces.
• Used a pretrained model fine-tuned for face-recognition.

Acknowledgements

• 3D-Text

facerecognition-in-arkit's Issues

From where do you show the image icon of the recognized person ?

Hi,

I went through the code but I couldnt make out how are you able to show the image icon of the person who's recognized like Boris with his image.

Where are those image being pulled from ?
and which part of code handles it.

Thanks.

add more info regarding data model n its training

Does not work on macOS Majove and Xcode 10

Basically, the libraries RxSwift and PKHUD don't work and the observer.onNext calls in your code flag as being a problem. There is no way to just pick up the code and get it running in Majove and Xcode 10 without doing a bit of work...

Feature request: Can we recognize multiple faces at once ?

Hello,
I am working on the demo where I have a similar requirement. But I have to recognize multiple person's faces at once. So here can we recognize multiple faces at once?
Thanks in advance.

Accidental Issue: Please Disregard :P

".mlmodel" file is missing

Hi All,

We have download and checked this project but we are getting compile error like ".mlmodel" file is missing.
Could you please update the latest source code to test this project? this help us to know more about this technology and move forward.

Thanks
Suresh

This model only recognize faces trained?

The training data is classified for each person.

1. Each person has over 300 augmented photos.
2. Each person is classified as one class.
3. Plus other photos not classified as a person above, rather classified just as other class.

In the example, there are 4 classes, 3 for each person, 1 for others.

Train with these photos would recognize only classified persons.
That is one of the reasons why model is not provided, because this model recognizes
only the three people.

build error

Do you know how to convert dlib_face_recognition_resnet_model_v1 to coreml?

Sorry for the super newbie question but this model generates "face features" that can be stored and compared later on.

The thing is I don't know what format it is so I can use the conversion tools provided by Apple to convert it to CoreML format.

TIA and thanks for sharing such a clean minimalistic code! ❤️

Maybe include the model so the thing can at least run

Run without having to build a new model from scratch

Pls update it into new version of swift 5

Pls update it to new version of swift 5