4 ms·

This is the second article in a short while I've seen about this FindFace app: https://news.ycombinator.com/item?id=11491264 https://news.ycombinator.com/item?i

This is the second article in a short while I've seen about this FindFace app: https://news.ycombinator.com/item?id=11491264 https://news.ycombinator.com/item?id=11491264 I am still amazed at how it is possible to match people in the way this app purports to do. Does anyone know of more accessible (to english speakers) code that can do roughly the same?

jessaustin 10y ago
Aren't FB and G+ pretty good at labeling people in uploaded pictures? Sure, they're working with a smaller universe and occasionally they need help from the user, but it seems fairly similar to this.
- egjerlow 10y ago
  You might be right, I hadn't really thought of those, mostly since I've never actually used the photo tag option. I'll experiment a little!
tyingq 10y ago
- This is open source, but I have no idea if it's anywhere near the FindFace capability: http://openbiometrics.org/ http://openbiometrics.org/ - Not open source, but it has a demo of face searching (http://www.faceplusplus.com/demo-search/ http://www.faceplusplus.com/demo-search/), and an api.
- asteadman 10y ago
  I'm not sure how leading edge the openbr stuff is, AFAICT (i'm not a domain expert or even novice), it's pretty primitive. If you were starting from scratch today, i'd say the openface project (https://cmusatyalab.github.io/openface/ https://cmusatyalab.github.io/openface/) is a better starting point. Having a DNN that maps images of faces into a point on a hypersphere seems eminently practical. I'm a bit surprised that it works, but having something like that is awesome.
  - tyingq 10y ago
    This post seems to indicate they aren't that far apart: http://bamos.github.io/2016/01/19/openface-0.2.0/ http://bamos.github.io/2016/01/19/openface-0.2.0/ I am though, in the same boat as you in this domain...not even a novice.
    - asteadman 10y ago
      That post is great! thanks for the link, i hadn't stumbled across it yet. From what I can tell from that link though, it is exactly as I suspected, and openface 0.2 significantly outperforms openbr for the LFW test set. (whereas openfaces 0.1 was largely equivalent). Moreover, I think that there is clearly room for drastic improvement with the openface architecture, either by tweaking the layout of the NN's, or simply throwing more data at it. It is not clear to me how openbr could be improved.

jessaustin 10y ago

Aren't FB and G+ pretty good at labeling people in uploaded pictures? Sure, they're working with a smaller universe and occasionally they need help from the user, but it seems fairly similar to this.

egjerlow 10y ago
You might be right, I hadn't really thought of those, mostly since I've never actually used the photo tag option. I'll experiment a little!

tyingq 10y ago

- This is open source, but I have no idea if it's anywhere near the FindFace capability: http://openbiometrics.org/ http://openbiometrics.org/ - Not open source, but it has a demo of face searching (http://www.faceplusplus.com/demo-search/ http://www.faceplusplus.com/demo-search/), and an api.

asteadman 10y ago
I'm not sure how leading edge the openbr stuff is, AFAICT (i'm not a domain expert or even novice), it's pretty primitive. If you were starting from scratch today, i'd say the openface project (https://cmusatyalab.github.io/openface/ https://cmusatyalab.github.io/openface/) is a better starting point. Having a DNN that maps images of faces into a point on a hypersphere seems eminently practical. I'm a bit surprised that it works, but having something like that is awesome.
- tyingq 10y ago
  This post seems to indicate they aren't that far apart: http://bamos.github.io/2016/01/19/openface-0.2.0/ http://bamos.github.io/2016/01/19/openface-0.2.0/ I am though, in the same boat as you in this domain...not even a novice.
  - asteadman 10y ago
    That post is great! thanks for the link, i hadn't stumbled across it yet. From what I can tell from that link though, it is exactly as I suspected, and openface 0.2 significantly outperforms openbr for the LFW test set. (whereas openfaces 0.1 was largely equivalent). Moreover, I think that there is clearly room for drastic improvement with the openface architecture, either by tweaking the layout of the NN's, or simply throwing more data at it. It is not clear to me how openbr could be improved.