A new service based on the NeoFace recognition system from NEC can provide data on who is visiting a store that can be used to fine tune marketing strategies.
NEC has quite a few AI-based projects and NeoFace is its state-of-the-art face recognition software. It has been highly rated by tests conducted by NIST as being fast and accurate.
You can license the software complete with an SDK to develop your own applications, but now NEC is using it to provide its own packaged service via its own "cloud" infrastructure. For $880 per month per store, the system will provide information about customers. All the store needs is a standard video camera and a PC; NEC's servers will do the rest.
The face recognition system detects faces in the video stream and can provide information such as sex and age. It doesn't use a face database to recognize customers in a wider context, but it does store data that enables the system to recognize a repeat visitor. The system will even recognize a repeat customer from another store using the same system.
It seems that the system processes the video data and provides the store computer with a stream of characteristic data rather than an identification, which means that the store cannot link the data to a particular shopper. This limits its usefulness to planning overall marketing campaigns, i.e. target women return shoppers on Monday morning, say.
Clearly the system is capable of providing much more information than NEC allows out of its servers and presumably it is only a matter of time before privacy considerations are ignored to build a more effective marketing system. You could object to this on the grounds that its another intrusion by "big brother". However, it could be presented as just an extension of the personal service provided by human shop assistants.
How spooky would it be to be greeted by a sales robot who already knows the sorts of things you like to buy?
Mapping is widely considered the most difficult perceptual problem in robotics, both from an algorithmic but also from a computational perspective. An innovative way around the problem relies on Rapyu [ ... ]