<div><br></div>I worked there 5 years ago. You absolutely can have databases that big, if you have the money to pay for the high end stuff. Keep in mind that this system was intended for government-level type use, and built to handle databases of that size. The company also employed many of the lead experts in the world on facial recognition algorithms. When I was there, the focus of research was way beyond how to match accurately, the focus had moved on to how to do it as fast as possible to get reasonable performance searching enormous databases.<div>
<br></div><div>We were de-duping databases such as Driver's Licenses for an entire nation, Immigration Visa applications, etc. We could also do things like search criminal databases (mugshots) that were enormous. Depending on the application, there would need to be a human to actually look at results to determine if it was an accurate match or not. </div>
<div><div><br><br><div class="gmail_quote">On Tue, Oct 18, 2011 at 11:35 AM, Yves Quemener <span dir="ltr"><<a href="mailto:quemener.yves@free.fr">quemener.yves@free.fr</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">On 10/18/2011 05:10 PM, Dan Wobser wrote:<br>
</div><div class="im">> Although with a high end software like the<br>
> one I worked on, we were dealing with databases of hundreds of thousands of<br>
> poor quality images and still getting over 90% accuracy.<br>
<br>
</div><div class="im">Wow, when was that ? I did not think you could have such a big database<br>
</div>nowadays with a reasonable error rate<br>
<div><div></div><div class="h5">_______________________________________________<br>
Discuss mailing list<br>
<a href="mailto:Discuss@lists.hackerspaces.org">Discuss@lists.hackerspaces.org</a><br>
<a href="http://lists.hackerspaces.org/mailman/listinfo/discuss" target="_blank">http://lists.hackerspaces.org/mailman/listinfo/discuss</a><br>
</div></div></blockquote></div><br></div></div>