Prepoznavanje lica – brzim putem strojnog učenja do povećanja učinkovitosti

Milan Bajić1, Matjaž Debevc2

mbajic@tvz.hr
1 Tehničko veleučilište Zagreb, Vrbik 8, Zagreb, Hrvatska
2 Univerza v Mariboru Fakulteta za elektrotehniko, računalništvo in informatiko, Smetanova ulica 17, Slovenija

Sažetak

Prepoznavanje lica opisuje računalno upravljano prepoznavanje ljudskih lica u digitalnom mediju, slici ili videu. Iako započeta u šezdesetim godinama prošlog stoljeća, novi val istraživanja prepoznavanja lica danas bilježi sve veći porast interesa djelomično i zbog sigurnosne primjene te smanjivanje mogućih sigurnosnih ugroza. U radu su dani smjerovi istraživanja u području, s naglaskom na korištene algoritme, skupove podataka za veću učinkovitost i pregled najčešće korištenih senzora. Jedna od novosti u području u posljednje vrijeme je korištenje različitih mehanizama strojnog učenja kao sredstva za povećanje učinkovitosti sustava za prepoznavanje lica. Predstavljanjem različitih autorskih pristupa dane su moguće smjernice za buduća istraživanja.
Ključne riječi: prepoznavanje lica, strojno učenje, algoritam, sigurnost

Face recognition – machine learning highway to efficiency

Abstract

Face recognition describes the automated computer managed human face recognition on digital media, image or video. Even though started in 1960’s, the hype surrounding face recognition nowadays is more increasing partially because of security applications it could provide and reduce possible security oriented threats. In this paper are given directions of the recent research in the field, with accent on algorithms used, data sets for higher efficiency and overview of sensors predominantly used. One of the concept that is introduced lately in the field is various types of machine learning as means for creating more efficient face recognition system. By presenting different approaches from authors the possible directions for the future research are given.
Keywords: face recognition, machine learning, algorithm, security


Patrik Ervells latest collection is ironically titled “Software”ADIDAS ORIGINALS TUBULAR SHADOW. In the notes from the presentation, Ervell says he was interested in developing tension between nostalgia and sci-fi,NIKE ZOOM WINFLO 3 and even included 90s cyberpunk amongst his influences. Ervell developed this point of view by creating an imaginary software company called Idegen software systems. He then re-appropriated the company’AIR JORDAN XXXI MENs logo in several of the collection’s garments. Elsewhere,MBT TARIKI MEN mohair coats with oversized lapels made an appearance alongside mock neck tees,NIKE AIR JORDAN RETRO 5 police-inspired ribbed sweaters,MBT FORA GTX WOMEN flight bombers, and polyurethane leather coats. The setting was also suitably dystopian, and the resulting ambiance was something akin to if the creators of Deutschland 83 and Hackers met up and decided to create a collection… We mean that as a compliment.Yesterday afternoon,JORDAN CP3 IX MEN artist JR spent the day working on his latest collaborative piece with Daniel Arsham in Greenpoint,NIKE AIR MAX 90 Brooklyn,MBT KIMONDO GTX MEN NYC. Connecting the lines between art, architecture, dance and theater, Daniel Arsham has been known to subvert existing architectural structures in unconventional,NIKE CLASSIC CORTEZ NYLON playful ways; confusing and confounding the expectations of space and form. Source: Street Art News .