Nowadays, one is not enough. The world wants more than one, so this is our motivation for work in order to meet the modern expectations. We have developed image recognition technology for one and more products to indicate a related content. We never ceased to improve it, but in order to keep pace with the leaders’ technology, we have to focus more on recognizing larger amount of products with our technology. The challenge is to make our API see a wider world and defeat a limited field of vision in the image recognition.
Now, our team is working on solution to identify shopping shelf tending to revolutionize merchandising system. The idea is, that using the app with IR, a user takes a photo of real in-store shelf with the products. Then, the technology recognizes all products captured in the photo, e.g. –chocolate, rice, beans, cookies, and returns content by displaying an information related to the appropriate one.
Each product is recalled by a suitable ID, which describes the proper product on the shelf. The position of the product is recalled in the form of four points attached to the concrete product. In the future, this function will allow developers to:
- count amount of products standing on the shelves,
- count a percentage of surface occupied by the products,
- verify a correct position of the product (product can stand straight, askew, upside down, etc.)
- take further merchandising initiatives with the product, e.g. . direct purchase of desired product – it requires advanced software and functions from the app.
We are working on a new algorithm, which differs from the algorithm which has been working in our API for over a year. The new one deals well with the multiple object recognition from one picture. (http://blog.recognize.im/index.php/2012/11/multiple-object-recognition-from-one-image-piece-of-cake/)
The old one returns a list of recognized products, but they have to be relatively large to be identified. We want to reach a sequence, which analyzes a larger image with a broader group of products to recognize (it means, that objects will be smaller in the picture). The consequence of recognizing more objects is, that the conversion is slower, but we will put a lot of effort into making this process faster.
To sum up, when we complete the new algorithm it will:
- recognize larger amount of objects,
- recognize smaller objects than can the present one,
- convert larger photos,
- speed up the recognition process to make productivity excellent.
We hope, that the visual merchandising will be a foothold of the future actions taken by the professional developers. We believe, that Our API has a potential to become an essential tool, which can enhance in-store experience and provide many benefits to the business and the ultimate consumer.