Patrick Grother, computer scientist, National Institute of Standards and Technology
18 December 2014 11:12 GMT

In the year when the FBI’s interstate photo system went into full operational capability, and when algorithms from two leading commercial developers demonstrated the potential of facial recognition - by finding U.S. fugitive Neil Stammer in Nepal and armed robber Pierre Martin in Chicago - activity expanded well beyond the criminal justice realm. 

In the civil sector there were open tenders in Australia, Morocco and the UK, together with several international eGate procurements.  As a NIST conference was told that eGate accuracy varies by passport issuer, the University of Bologna commenced testing passport photo compliance software.  

Government activity was mirrored in the private sector, where Google revamped Android's face authentication, Intel engaged Cognitec on mobile device access, Dropbox acquired KBVT developers to index faces and other content, and Facebook published its use of massive social media sets to develop its DeepFace algorithm. 

Driven by the ubiquity of cameras, the next 12 months will see both formalized practices on the acceptable public use of face recognition technology and an expanded marketplace of algorithms.  In turn this will be accompanied by more accessible and more robust testing of those algorithms, particularly for the identification task that remains unaddressed by the popular LFW benchmark.

The year ahead will see better characterization of how absolute age and also ageing affect accuracy.  Similarly we’ll get better insight into how, why and when algorithms outperform human experts (and vice versa).  Finally 2015, will see expanded transaction volumes, in everything from personal authentication to eGates, to fraud detection and criminal investigation. 

Return to the index ….. Or read the next review