Hacker News new | ask | show | jobs
by wpietri 2987 days ago
How good would you say the tech is? E.g., if Facebook put cameras outside a football stadium with 60k people, how many people would they correctly recognize? And how many false positives would be generated?

I ask because I'm wondering how much China's announcement here is a stunt that goes well beyond the state of the art.
2 comments
bsenftner 2987 days ago
The tech is there, it works. The current hurdle is getting understanding between the practical limitations of modern computers, the resolutions and image characteristics of various cameras. Trying to use a 2K or 4K video stream simply overwhelms the current generation of processors in the systems 'normal' organizations can afford. Plus those video streams are often up-scaled and are lower resolution in their hardware, very wide angled, and placed poorly for FR. Basically, the tech is there, but the practical aspects of using FR with typical financial and logistical constraints is not there yet. This practical aspect is just the normal learning curve.

Facebook's FR is nowhere but FB, forget it. There is a range of false positive rates, depending upon how an FR system is employed. The best resource for the state of the art is probably this: https://nvlpubs.nist.gov/nistpubs/ir/2017/NIST.IR.8173.pdf (FYI, my work is vendor "L")

link
jhartmann 2987 days ago
Depends on the cameras, but I would imagine Facebook and Google's tech + data to be able to positively identify 99% or more of the people. Especially with multiple shots per person. Accuracy might be as high as 99.9% if you have drones go over the crowds, and have high res cameras at chokepoints. Correlate this with cell phone data and probably they have 100% accuracy. There is a reason Google Fi exists.

I say all this as a happy Google User and a developer of Face Recognition Technology. Those in the know realized privacy has been dead for awhile now. Generally computers can do most anything we can do in say 200 ms or less with the proper training data, and they generally do it better than humans. The giants have all the data they need, they just need to spend time annotating training sets and running analysis over larger and larger sets until they reach super human performance.

link