Researchers Develop Fast Algorithm for Distinguishing Weighted Goods in Supermarkets

Researchers Develop Fast Algorithm for Distinguishing Weighted Goods in Supermarkets


A team of researchers from Skoltech and other institutions has introduced a new algorithm capable of quickly distinguishing weighted goods in supermarkets. Unlike existing systems, this algorithm boasts the ability to expedite neural network training when new types of produce arrive. Their work has been published in the IEEE Access journal.

The Need for Efficient Weighted Goods Detection

Supermarkets and grocery stores often struggle with quickly and accurately identifying different types of weighted goods, which can lead to longer checkout times and frustration among customers. Existing systems typically use traditional computer vision techniques, which can be slow and limited. This creates a pressing need for a more efficient method.

The New Algorithm’s Approach

The research team aimed to address this issue by developing a faster algorithm that relies on neural network training. They created a unique dataset consisting of various fruits and vegetables, including produce with deformities or damage, to train the algorithm. The researchers then trained the algorithm using the newly created dataset, enabling it to quickly and accurately recognize different types of weighted goods.

Benefits of the Algorithm

While existing systems require manual labeling and extensive training to detect new produce types, the new algorithm eliminates this cumbersome process. The researchers discovered that the algorithm’s performance remained consistent even when faced with produce it had not encountered during training. This means that supermarkets and grocery stores can introduce new products without significant delays to the checkout process.

Promising Results

The researchers tested the algorithm on a prototype system that included an off-the-shelf camera and a small personal computer. They observed that the algorithm identified various types of fruits and vegetables with impressive accuracy and speed. This demonstrated the feasibility of implementing the algorithm in real supermarket checkout systems.

Future Directions

Moving forward, the researchers plan to further refine the algorithm and investigate potential commercial applications. They aim to collaborate with industry partners to integrate the algorithm into existing checkout systems, ultimately improving the shopping experience for customers.


The development of this fast algorithm for distinguishing weighted goods in supermarkets brings hope for more efficient checkout processes. By utilizing neural network training, this algorithm eliminates the need for labor-intensive manual labeling and allows supermarkets to introduce new produce without delays. The results have shown promising accuracy and speed, making it a notable advancement in the field. As researchers continue to refine and explore commercial applications, the algorithm has the potential to revolutionize the way supermarkets identify and process weighted goods.

More from this stream