Can anyone explain that to me, how it works once the picture is in the database. Then what? You walk on a street and a cctv camera captures a picture of you and it can be compared later with other pictures in a db, based on some patterns? How about the resolution or noise? Any tech guys here?
I am no expert or would even call myself knowledgeable. However, I did write up some technical documents about a year back. The software at that time is very good at recognising facial structure and can identify gender without any previous record of the subject, even from odd angels it can achieve high recognition from a database fairly rapidly. The only downfall of that system (and apparently most facial recognition of the time) is it has trouble with emotion; if you had on your poker face for the database picture it wouldn't recognise you if you are smiling even from a perfect angel.
Resolution and noise seem only to have an effect on how high the confidence the system can place in the identification, a closed system as in a university would probably be able to identify above 90% since the pool will be smaller. Using a system like this to find a person who you say know is in down town London and checking all the cctv, I don't think is possible, I don't think the technology is there yet.
As I said, I am not an expert and this was only knowledge of one system (a very expensive one). So if someone has more accurate information I yield to them.
Man, that is creepy. Angle, facial expression and other will come with time. Eg. Facebook. Some people are forced to submit an official ID in order to keep the account. So there is already "an official facial" expression that can be linked to other pictures that people submit, expressing their emotions.
All these things combined, as they appear to me, seem that for instance, the facebook network is an elaborated database for harvesting data of individuals. Imagine what happens when you just put other networks (not officially connected to each other) into one vast dataset. And it is for free, since people are not forced to submit personal information.
view the rest of the comments →
puag ago
Can anyone explain that to me, how it works once the picture is in the database. Then what? You walk on a street and a cctv camera captures a picture of you and it can be compared later with other pictures in a db, based on some patterns? How about the resolution or noise? Any tech guys here?
OWNtheNWO ago
The military has been using handheld devices in Iraq for over a decade that can read your biometric faceprint from a hundred yards away.
nearly-evil ago
I am no expert or would even call myself knowledgeable. However, I did write up some technical documents about a year back. The software at that time is very good at recognising facial structure and can identify gender without any previous record of the subject, even from odd angels it can achieve high recognition from a database fairly rapidly. The only downfall of that system (and apparently most facial recognition of the time) is it has trouble with emotion; if you had on your poker face for the database picture it wouldn't recognise you if you are smiling even from a perfect angel.
Resolution and noise seem only to have an effect on how high the confidence the system can place in the identification, a closed system as in a university would probably be able to identify above 90% since the pool will be smaller. Using a system like this to find a person who you say know is in down town London and checking all the cctv, I don't think is possible, I don't think the technology is there yet.
As I said, I am not an expert and this was only knowledge of one system (a very expensive one). So if someone has more accurate information I yield to them.
puag ago
Man, that is creepy. Angle, facial expression and other will come with time. Eg. Facebook. Some people are forced to submit an official ID in order to keep the account. So there is already "an official facial" expression that can be linked to other pictures that people submit, expressing their emotions.
All these things combined, as they appear to me, seem that for instance, the facebook network is an elaborated database for harvesting data of individuals. Imagine what happens when you just put other networks (not officially connected to each other) into one vast dataset. And it is for free, since people are not forced to submit personal information.