face-api.js Liveness Detection

Hi,

I am working on an internal room check-in system using this API. The detection speed and accuracy looks great. But there is a problem that people can use pictures on their phone or printed paper to unlock the door. I am wondering if there is any liveness detection (anti-spoofing) model we can integrate to make this API better. Thanks

Rio

Feb 20 '19 21:02 GL-RuiLa

Also interested in liveness detection.

One aspect of that could be blink detection, see #176.

Feb 22 '19 18:02 nathany

Also interested! Something like that! http://cvlab.cse.msu.edu/project-face-anti.html

Mar 01 '19 23:03 leandrojo

An article on one type of liveness detection using a convolutional neural network: https://www.pyimagesearch.com/2019/03/11/liveness-detection-with-opencv/

Mar 12 '19 18:03 nathany

Also Interested!

Apr 16 '19 16:04 guinaldi

I would love to see Vincent @justadudewhohacks implements some kind like this https://github.com/richmondu/libfaceid

Jun 02 '19 09:06 Bajajar

This is likely my #1 request for this fantastic library.

Sep 13 '19 12:09 ScottDellinger

I have converted liveness keras model from https://www.pyimagesearch.com/2019/03/11/liveness-detection-with-opencv/ into tensorflowjs json model. anyone know how can i use this converted model in this library?

live

Oct 11 '19 03:10 ariffromeo

New open source solution: https://github.com/SoftwareGift/FeatherNets_Face-Anti-spoofing-Attack-Detection-Challenge-CVPR2019

international download link model

Testing code

Output:

./images/fake.jpg: FAKE!
./images/real.jpg: REAL

Nov 05 '19 18:11 msvargas

An article on one type of liveness detection using a convolutional neural network: https://www.pyimagesearch.com/2019/03/11/liveness-detection-with-opencv/

I tried that but the accuracy of the model is very poor.

Jan 22 '20 11:01 ankitnamdeo34

Landmarks, specifically for eyes, changes in difference of distance between upper eye points and lower eyes points

We tried to use the same for fatigue detection but sadly eyes closed, it wasn't perfoming well for Indians faces so had to move to a different solution.

Summarizing, record a video run face-api notice how distances are changing, disproportionately compared to other landmarks, they are alive if not it's something static which may be moving around as a whole

Feb 11 '20 00:02 no-1ne

Hello there,

Just wanted to suggest a direction on how to do this.

There's a company which I won't disclose the name, who does the liveness detection based on the realness of the skin of the person.

While capturing the video input, several screens of different colors are quickly displayed on the screen. The algorithm analyses the reflection of light on the skin and determines if it's real or not.

Chey