Computer Vision News - November 2016

Owing to the need of a big screen to see all of the products on the shelf, Yodigram’s customers tend to work with tablets instead of using mobile phones. The best tablet screen featuring now an 8-megapixel camera, it does not work well with shading, etc. If you have a strange light, you see nothing. That’s why it was difficult to have actual OCR for the recognition of the product and that’s why they gave up OCR. Now they are trying to add more functionality. For example, if you didn’t take a good picture and the photo is bad quality or not sharp enough, they are adding a function that alerts the user on the spot so that he can take another picture when he is still in the shop. The unicity of app comes from that it provides instant detection: when you have almost 60 or 70 products in the same photo, you can detect the exact products in the photo with the same accuracy as if you used the barcode. The one and a half minutes that take the complete calculation for one picture is short enough to enable the user to watch the image and find the solution on the spot: what exactly is on the shelves, what needs to be restocked, why things are arranged as they are, and so on. From the image, product display can be compared to that of competing brands. Yodigram gets pictures from other brands too, so that user can compare his or her products’ display with that of directly competing products. It also offers metrics, for example, showing what products can be viewed at eye level. That might lead to expand the eye level exposure and improve marketing. 68 Computer Vision News Application Application Original shelf Debug