Article

2019-07-19 Science & Technology

Why beards are holding you up at the airport: Trendy face fuzz blamed for long queues and even flight delays

Archived material is restricted to Rantburg regulars and members. If you need access email fred.pruitt=at=gmail.com with your nick to be added to the members list. There is no charge to join Rantburg as a member.

Posted by Skidmark 2019-07-19 10:01|| || Front Page|| [4 views ] Top

#1 404 not found

Posted by warthogswife 2019-07-19 11:39|| 2019-07-19 11:39|| Front Page Top

#2 Fixed: Dailymail

Posted by Frank G 2019-07-19 11:49|| 2019-07-19 11:49|| Front Page Top

#3 Gosh, it's almost as if facial recognition is an over-hyped, lame-ass technology that shouldn't be let loose outside the lab. Prolly good money in sales, though. And I say that as a machine-learning fanboi.

Posted by SteveS 2019-07-19 12:29|| 2019-07-19 12:29|| Front Page Top

#4 Actually facial recognition is very easy to do.

Most of the algorithms are now open source. The quality of the predictions/outputs of these algos depends mainly on the quality of the data, ie the # and variety of the facial image inputs.

The field is just an offshoot of computer graphics, with relies on solutions to a set of math problems involving topology, parametric surfaces and other branches of geometry that have been well understood by the industry for decades.

All that was missing was a reliably large data set of facial images.

If the input variables now include billions, or even trillions, of facial images, then it's a certainty that we will get to 99.9% reliable facial recognition for most of the population in the near term.

Posted by Lex 2019-07-19 12:39|| 2019-07-19 12:39|| Front Page Top

#5 I didn't say it was hard to do. I'm questioning the accuracy of the implementations. There is no question the algorithms are well known, but you still have to train the model. And there, the devil is in the details.

There is a big difference between finding a face in an image and recognizing a specific face. While we have lots of data in terms of pictures of faces, we don't have lots of data for a specific face. And lots of data is the key for training neural networks.

Posted by SteveS 2019-07-19 13:31|| 2019-07-19 13:31|| Front Page Top

#6 If they use a 3D camera system it's easy to auto-clip most of the beard.

Posted by 3dc 2019-07-19 16:32|| 2019-07-19 16:32|| Front Page Top

#7 SteveS - I am indeed saying it is easy to train models that ACCURATELY predict the face owner's identity. Especially if the prediction need not be processed in near-real time.

This is entirely a data quantity problem, not a hard data science or data engineering problem.

Posted by Lex 2019-07-19 16:34|| 2019-07-19 16:34|| Front Page Top

#8 Artificial problem.
With so many touch surfaces DNA cannot be disguised or denied. Blockchain the Genome identity.

Posted by Skidmark 2019-07-19 19:00|| 2019-07-19 19:00|| Front Page Top

23:04 Clolutch Shaving8167
23:00 Snavimble Bucket1794
22:52 DarthVader
22:50 M. Murcek
22:22 Whiskey Mike
20:56 trailing wife
20:51 trailing wife
20:49 Glenmore
20:42 Glenmore
20:41 ed in texas
20:40 ed in texas
20:39 trailing wife
20:38 jpal
20:30 Omorong Black8889
20:28 Mrs. Davis
20:00 gorb
19:24 DarthVader
19:24 Besoeker
19:22 3dc
19:19 Silentbrick
19:15 Besoeker
19:14 Alaska Paul
19:13 Bobby
19:12 Alaska Paul