GADAI

Automatic retail shelf management using visual information and AI.

Description

The aim of the project is to optimise the shelf management process of retail shop selling plans. This is an activity that allows the Store Account Manager to get a precise picture of the situation. The solution allows the following information to be recorded on the shelf for each individual product: number of items displayed on the shelf, how much space the product occupies, shelf number, absence of product on the shelf, retail price and promotions. The analysis platform developed involves the use of image capture devices and a reporting dashboard. Data acquisition is performed by a camera positioned on the shelf or by a camera of a smartphone/tablet used by the shop staff. Finally, a cloud-based machine learning and image processing algorithm was developed to analyse the products on the shelves.

Result to be enhanced

At this moment, several tests of the application developed have been carried out. These tests have been carried out both in vitro, using images previously captured and loaded from the device’s memory, and in vivo, by taking photos at points of sale representing the use case. All the tests performed demonstrate the robustness of the application, which is able to provide a response to the user in a short time with an accuracy of over 90%. Finally, a high-performance demonstrator was developed and presented to potential customers.

Why is it important?

In recent times AI, and in particular machine learning, has considerably improved the ability to automatically recognise objects from images and videos. Today, the task performed by store accounts of a company producing goods distributed in retail chains (e.g. supermarkets), is to check the expository situation in the sales plan. At present, the detection of the shelf status is carried out manually by confirming or entering the data of each individual product re-presented in the system from the previous detection. This activity results in a loss of time that is prone to errors. The solution developed allows the Store Account Manager to have a detailed and real-time picture of the situation, saving time and having everything constantly under control.