As is well known, the position of a product on a shelf, the number of its instances exposed in front of a client and the proportions of several products play a significant role in the sales process and have a strong impact on revenues. Over two years ago, the Recognize.im Team released a specialized algorithmic engine to support this task – the Shelf Mode, which allows analyzing the shelf contents. The time has come to empower it with dedicated hardware.
Since the recognition results depend strongly on the quality of query pictures, the Recognize.im Team came up with the idea of replacing an inaccurate human holding a random smartphone with a dedicated device intended for installation in front of shelves. Such an appliance could provide much better data for image recognition algorithms by taking homogeneous pictures in a recurrent way, thus leading to significant improvement of recognition results. The device – the Steady Shelf Scanner – is energy-saving, compact, easy to assemble, robust and, above all else – inexpensive.
The prototype is a low-power mini-computer with a tiny on-board camera, capable of taking a high-quality picture and sending it via the Internet for further analysis. In order to make the query picture capturing process even more refined, the prototype has been equipped with the following extra features:
– a motion/presence sensor, which detects whether there is anyone moving/standing between the camera and the shelf when a picture is being taken;
– a light sensor, which calibrates the camera parameters (ISO, white balance, etc.) and helps save energy by putting the device in standby mode when it is dark (during the night).
The first tests in real-world conditions proved that using the device increases the reliability of the analysis results. The dedicated camera calibrated for this particular purpose and combined with smart sensors ensures high quality of every single query picture. Image framing can be optimized and any distorting effects of camera movement are eliminated.
The camera itself does not have to be fixed to the device. Since it is tiny, it can be installed separately wherever it fits, or it can even be embedded into the shelf bar, where price labels are usually placed. In a configuration where two shelves stand in front of each other, a camera located on one shelf is used to check the contents of the other one. There are several advantages of this solution:
– the camera stays hidden and physically secured;
– the Scanner can be located in a place that is beyond the reach of the store’s visitors;
– the camera is placed exactly in front of the shelf being analysed.
The prototype of the Steady Shelf Scanner connected to the Recognize.im Shelf mode returned very promising results already after the first tests under real conditions. With this outcome and the knowledge on both the prototype and the Image Recognition Technology, we are ready to install and arrange the entire pieces of described systems in real facilities in order to examine the concept of the Shelf Contents Analysis System in detail, improve it and prepare both the devices and the software for mass production and installation.
Having access to a network of such devices using the Shelf Mode for products recognition, one can actually generate a real-time and precise shelf contents analysis with a single mouse click. The variety of reports and conclusions that can be achieved using shelf analysis is limited only by the imagination of the system’s Owner.
We are ready to install the proposed solution in your own facility to show you how it works and to let you use it. We would be happy to learn what your needs and expectations are in order to make the presented solution more customizable, practical and businesslike.
We hope that the visual merchandising will be a foothold for the future actions taken by professional developers. We believe that our API combined with the Steady Shelf Scanner has the potential to become an essential tool that can enhance in-store experience and provide many benefits to the businesses and the ultimate consumer.