Blippar's computer vision (artificial intelligence) accuracies have been independently verified with publicly available data sets

The Department of Engineering at the University of Cambridge has done an independent evaluation of the accuracy of Blippar's computer vision capabilities in areas of Car Recognition, Face Recognition and Face Verification.

The first section of the report evaluated Blippar’s car recognition technology, assessing its ability to distinguish between different makes and models of cars. The software was evaluated alongside Sighthound, a leading publicly available car recognition technology that officially claims to cover “almost all popular models in North America.”

The analysis used three separate datasets of various makes and models in the US car market – Blippar’s own database of US cars, an external set gathered from online marketplace eBay, and the world’s largest public collection of car images, compiled by the Chinese University of Hong Kong. On each dataset, Blippar’s performance was noted ‘more accurate’ and ‘superior’ to the Sighthound API.

The second section of the report analysed Blippar’s facial recognition and verification technologies. For this section, the researchers created a quantitative and qualitative analysis of Blippar’s computer vision software.

The team of researchers assessed Blippar’s face verification technology, using the external LFW dataset, which includes 13,233 images of celebrities taken from Yahoo News. Blippar’s technology matched faces on this database with an accuracy of 99.67%.

The researchers compared Blippar's recognition to those of several other major players for example they have compared Blippar's face recognition to Amazon and Microsoft's APIs. They also mentioned that Blippar's face recognition had a high accuracy performance under varying head pose/rotation, scale and illumination.

The conclusion of the report was " On the challenging evaluated datasets, Blippar’s car and face recognition technologies demonstrated superior performance to the other tested leading competitors whose technologies are publicly verifiable.”

You can read the full report here.