Scholarlink

Demand for efficient and accurate billing systems continues to grow as retail environments become increasingly dynamic. This paper presents an advanced vision-based billing system leveraging You Only Look Once (YOLO11), a state-of-the-art object detection mode, in order to enhance the checkout process by enabling precise product detection, counting, bundling and invoice generation. With a robust dataset from Roboflow and a carefully optimized training pipeline, the model achieves superior accuracy (mAP@[0.5:0.95] of 94.84%) and computational efficiency over YOLOv9-t(mAP@[0.5:0.95] of 93.04%) and YOLOv10-n (mAP@[0.5:0.95] of 92.36%), making it well-suited for cluttered retail scenarios. A co-occurrence-based bundling module is introduced, which identifies meaningful product associations from transactional data, facilitating personalized marketing strategies and cross-selling opportunities. Moreover, the model is integrated with a web application to demonstrate its practical usage. This integration of high-speed object detection with actionable customer insights offers a lightweight yet powerful solution for retail automation.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Vision-Based Automated Billing System Using YOLO11: Integrating Object Detection and Product Bundling for Retail Optimization

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sameer S. Mansur,
  • Ishan G. Kulkarni,
  • Vedaant Mathreja,
  • Saisatwik A. Madalageri,
  • Channabasappa Muttal,
  • Sneha Varur

摘要

Demand for efficient and accurate billing systems continues to grow as retail environments become increasingly dynamic. This paper presents an advanced vision-based billing system leveraging You Only Look Once (YOLO11), a state-of-the-art object detection mode, in order to enhance the checkout process by enabling precise product detection, counting, bundling and invoice generation. With a robust dataset from Roboflow and a carefully optimized training pipeline, the model achieves superior accuracy (mAP@[0.5:0.95] of 94.84%) and computational efficiency over YOLOv9-t(mAP@[0.5:0.95] of 93.04%) and YOLOv10-n (mAP@[0.5:0.95] of 92.36%), making it well-suited for cluttered retail scenarios. A co-occurrence-based bundling module is introduced, which identifies meaningful product associations from transactional data, facilitating personalized marketing strategies and cross-selling opportunities. Moreover, the model is integrated with a web application to demonstrate its practical usage. This integration of high-speed object detection with actionable customer insights offers a lightweight yet powerful solution for retail automation.