3 ms·

Unfortunately, Apple being Apple - doesn't expose those new sensors in any APIs.

wlesieutre 9y ago
And thank god for that. The face recognition data goes in the secure enclave and stays there. The last thing I need is Snapchat to say "We need access to the facial scan for new video filters!" and then using it to measure microexpressions in response to ad displays.
- ceejayoz 9y ago
  They could conceivably provide APIs that refuse to give detailed scans when a face is detected.
  - wlesieutre 9y ago
    Can't you do that with the regular camera? The only reason you need the grid projector is for getting a detailed 3D model. Haven't checked, but I assume that (like TouchID) they do provide APIs to use it for authentication, but they refuse to give any more data than that.
    - ceejayoz 9y ago
      > The only reason you need the grid projector is for getting a detailed 3D model. That's what I want it for. TouchID isn't very useful outside of fingerprinting. The dot projector I could see a wide variety of really neat, innovative uses by third-party devs if it can be made accessible.
      - wlesieutre 9y ago
        Given how explicit Federighi has been about the FaceID data staying on-device in the secure enclave, with not even Apple having access to it, I think it's very unlikely they'll change their stance on that. The moment 3rd party devs have access to the scanner, it'd be siphoned off by somebody's ad framework and then sold for biometric ad targeting in nearby stores or god knows what else.
        ceejayoz 9y ago
        Hence my saying: > They could conceivably provide APIs that refuse to give detailed scans when a face is detected.
        wlesieutre 9y ago
        Oh I see, I'd read that as generating some kind of simplified "generic face" model to use for face features without biometric identifiers, but you're talking about using the grid projector for more generic stuff like people had done with Kinekt sensors?
        ceejayoz 9y ago
        Yes, exactly. 3D scans of objects, room mapping, robot brain...
        wlesieutre 9y ago
        Pretty clunky with the sensor on the front, but I agree it'd be cool. For room mapping it seems like ARKit should cover a lot of use cases. The tape measure demos I've seen were solid. But IIRC ARKit has a lot of stuff built around identifying flat surfaces like that, it can't help you so much with arbitrary objects.
bpicolo 9y ago
They demo'd Snapchat with improved face tracking, though?

And thank god for that. The face recognition data goes in the secure enclave and stays there. The last thing I need is Snapchat to say "We need access to the facial scan for new video filters!" and then using it to measure microexpressions in response to ad displays.

ceejayoz 9y ago
They could conceivably provide APIs that refuse to give detailed scans when a face is detected.
- wlesieutre 9y ago
  Can't you do that with the regular camera? The only reason you need the grid projector is for getting a detailed 3D model. Haven't checked, but I assume that (like TouchID) they do provide APIs to use it for authentication, but they refuse to give any more data than that.
  - ceejayoz 9y ago
    > The only reason you need the grid projector is for getting a detailed 3D model. That's what I want it for. TouchID isn't very useful outside of fingerprinting. The dot projector I could see a wide variety of really neat, innovative uses by third-party devs if it can be made accessible.
    - wlesieutre 9y ago
      Given how explicit Federighi has been about the FaceID data staying on-device in the secure enclave, with not even Apple having access to it, I think it's very unlikely they'll change their stance on that. The moment 3rd party devs have access to the scanner, it'd be siphoned off by somebody's ad framework and then sold for biometric ad targeting in nearby stores or god knows what else.
      - ceejayoz 9y ago
        Hence my saying: > They could conceivably provide APIs that refuse to give detailed scans when a face is detected.
        wlesieutre 9y ago
        Oh I see, I'd read that as generating some kind of simplified "generic face" model to use for face features without biometric identifiers, but you're talking about using the grid projector for more generic stuff like people had done with Kinekt sensors?
        ceejayoz 9y ago
        Yes, exactly. 3D scans of objects, room mapping, robot brain...
        wlesieutre 9y ago
        Pretty clunky with the sensor on the front, but I agree it'd be cool. For room mapping it seems like ARKit should cover a lot of use cases. The tape measure demos I've seen were solid. But IIRC ARKit has a lot of stuff built around identifying flat surfaces like that, it can't help you so much with arbitrary objects.

bpicolo 9y ago

They demo'd Snapchat with improved face tracking, though?