INTELLIGENT SYSTEMS AND TECHNOLOGIES

BARdger: AI-boosted configurable barcode scanning system

V. A. Karyakina^a, V. A. Karnaushko^ab, D. P. Nikolaev^ab, V. V. Arlazarov^ab, K. B. Bulatov^ab, P. V. Bezmaternykh<sup>ab

a</sup> Smart Engines Service, Moscow
^b Federal Research Center "Computer Science and Control" of Russian Academy of Sciences, Moscow

Abstract: We present BARdger, a barcode scanning system that combines latest machine learning achievements, computationally efficient image processing techniques, and flexible configuration capabilities. It addresses actual challenges in barcode recognition, including damaged codes reading, variable lighting conditions, and diverse symbology support. The system architecture integrates lightweight neural networks optimized for parallel and high-performance processing with well-established procedures for reliable bar-code reading. Its configurable design enables users to select AI models based on performance requirements, customize preprocessing stages without system recompilation, adjust recognition parameters on-the-fly, and execute it in a WebAssembly environment. We illustrate several notable system configurations and benchmark the BARdger over the BarBeR framework. The system provides a foundation for developing and integrating novel methods and algorithms in the field of barcode recognition, including challenges such as reading irregular matrix codes or those with significant blurring.

Keywords: barcode scanning, image processing, deep learning.

Language: English

DOI: 10.14357/20718632250401

Bibliographic databases:

•