What's the effective way to tag aligned face and to recognize them quickly?

[xyx2011]

Provide Your Network Information

1. Where is the location of your server? (i.e. City, or In/Out China)
   in China, Shanxi

2. Which cloud platform(AliYun/Qcloud/DigitalOcean/etc) are you using?

Self-hosted

Expected behavior

There is a API to take aligned faces and tag them (face A = 'bob', face B = 'Steve'), then when an unknown face come in from webcam, we can use this API to label who that person is under 1s.

Actual behavior

no such an API

Steps to reproduce the behavior (and fixes, if any)

simple solution: KNN (but doesn't scale well when there are 1000+ faces)
better solution: SVM (still doesn't scale well when there are 1000+ faces)

Does anyone have other ideas?

[huan]

Hi @xyx2011 ,

Thanks for this feature request.

From the function view, this request is more like to be an enhancement to the face-blinder module, you can learn more about it at https://github.com/zixia/face-blinder

From the performance view, I'd like to suggest you have a look at K-D Tree, you can learn more about it at https://en.wikipedia.org/wiki/K-d_tree