25.10.2019

foto Petr Bravenec

Petr Bravenec
Twitter: @BravenecPetr
+420 777 566 384
petr.bravenec@hobrasoft.cz

Historical review

Advances in software development and the availability of poweful computers allows us to do things that were hard to imagine before. In my first job I programmed wage agenda on SMEP computers (PDP-11 alternative). I used the Redap and Fortran programming languages. I remember a funny story: The boss came and says to my team: There should be the duplicity check of personal number before a new person is created! At the moment when boss leaved the room, we started to laugh. I could imagine how to solve the problem, but it was very hard to achieve that with recent computers ant it could take months of work.

In these days such check can be made automatically in database engine with no developer's effort. The unique personal number (primary key) would be part of data properties and it would not be possible to make a record with duplicate personal number.

Already at that time computers exist in sci-fi literature such was famouse HAL-9000. HAL-9000 observed the world with camera and it would be easy for him to turn-on my computer and make me a cofee when I arrived to my office. HAL-9000 recognized the crew members by their faces. It was clear sci-fi in 1990. Nobody could imagine how to make it. Such request would be seen automatically as a joke in 1990.

Well, what has happened since then? Ten years ago I experimented with OpenCV and I tried to find faces on image. I could recognize and frame the face, but it was impossible to idetify the person. Nobody known how to make it and no cheap, powerful and low-comsumption computer existed in that days.

Face recognition with Nvidia Jetson Nano

Last week I bought Nvidia Jetson Nano. When I connected camera, it was mather of two afternoons to make a simple face recognition application. I loaded images of all office members and our children. Especially children love the application. In other direction - the application does not love children, it was mistaken very often. But adult faces the aplication recognizes very easy and with high precision. Whole computer has consumption less the 10 watts. I can imagine that such computer could run 24/7 in our office and it could perform some tasks based on face and gesture recognition:

  • gesture "I'm working with jackhammer" should wake-up person's computer, light her desk lamp and make a record to attendance book,
  • gesture "Bye bye" should suspend the computer, make a record to attendance book and turn-on the alarm system when last person leaved,
  • it should not react to children, only the alarm should be supressed,
  • other persons with office crew member escort should be ignored,
  • all unexpected visitors should be shot in their head.

See how the face recognition works:

Thanks to face recognition library it was very easy to make such application. Of course, in real word the application requires much more features: connection to recognized persons database, connection to information system, it could react specifically to recognized person etc. But that is only a routine job which is made be millions developers every day.

Hobrasoft s.r.o. | Contact