The facial recognition subsystem was created as a custom order for integration in an offline advertising project. The subsystem’s primary task is to document, with given precision, the glances at the advert screen from the people who appear in the view of the web camera installed above the digital screen, and to correlate this data with the ad stream in order to receive data about ad materials that are actually viewed or ignored by the target audience. The purpose of the subsystem is not only assessment of actual effectiveness of this type of advertising in general, but also of certain materials separately.
Our company’s efforts have developed a hardware-software comprehensive set that includes a portable computer, a web camera, as well as software, that performs a set task. This hardware-software comprehensive system performs the following key functions: determination of presence of a human face in a snapshot; details and calculation of direction of a glance on discovered faces; tagging the previously discovered faces for measuring the actual duration of a glance at a media vehicle; determination, wherever possible, of gender and age characteristics for issuance of recommendations on targeting of advertising materials; transfer of collected data to the server and correlation with the real media stream in order to calculate the data ownership of each specific advertiser.
These technical screenshots are made with the involvement of the company employees in the conditions of imitation of real usage in terms of lighting and generic geometric installation parameters. The shots are handheld, which doesn’t interfere with images. Within the set’s normal functions, neither operators nor advertisers have access to images of final persons and similar technical images.