search

FaceONNX / FaceONNX

Face recognition and analytics library based on deep neural networks and ONNX runtime
MIT License
179 stars 33 forks source link

How well does faceONNX resist face spoofing? #2

Closed ADD-eNavarro closed 2 years ago

ADD-eNavarro commented 3 years ago

Hello. First things first, thanks for your work, looks awesome. I was wondering if you have considered security in your implementation, or are planning to include some, to prevent face spoofing.

asiryan commented 3 years ago

Hey! Thanks for the feedback. As for anti-spoofing, this technology requires special sensors and cameras. At the same time, there are several fundamentally different approaches to solving the problem of identity and security verification. For these reasons, the corresponding neural network will not be available within the FaceONNX library.

ADD-eNavarro commented 3 years ago

Thanks for such a fast response. Well, actually that's exactly what I'm after, a PAD system that requires no particular hardware. I've been looking into MiDaS for monocular depth perception, in the belief that pairing face detection with depth could achieve this result, for example through face landmark points depth comparison, since a presentation attack would be flat (unless it's a mask). Anyway, even without antispoofing, I consider your work is great. Thank you again for it.

asiryan commented 3 years ago

Thanks for such a fast response. Well, actually that's exactly what I'm after, a PAD system that requires no particular hardware. I've been looking into MiDaS for monocular depth perception, in the belief that pairing face detection with depth could achieve this result, for example through face landmark points depth comparison, since a presentation attack would be flat (unless it's a mask). Anyway, even without antispoofing, I consider your work is great. Thank you again for it.

You are moving in the right direction, you need depth maps.

Best regards

asiryan commented 2 years ago

Hi @ADD-eNavarro! I want to present new feature of FaceONNX -> FaceDepthClassifier. It can be used for 3D antispoofing algorithm. Check this out.

ADD-eNavarro commented 2 years ago

That is AWESOME! I'll have a look and try the examples, thank you so much!

asiryan commented 2 years ago

That is AWESOME! I'll have a look and try the examples, thank you so much!

I'm thinking of posting a repository with a full-fledged Face ID system based on Intel RealSense cameras. This will allow you to create your own implementations of Face ID systems using my FaceONNX library.

ADD-eNavarro commented 2 years ago

Well, that would be nice for sure. But then you link your system to some particular hardware, which I wouldn't but of course it's up to you. I'll be lloking into the face depth example this weekend. Can't wait!

asiryan commented 2 years ago

@ADD-eNavarro Work on the new components of the library has been completed. The current version, 2.0.0.1, includes all the necessary anti-spoofing tools. Also follow my new repository, I will post the FaceID in it.