.png){kind=icon}
.png){kind=icon}
Return
On-shelf availability (OSA) detection using machine learning approach
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-026-21323-5.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Ensuring on-shelf availability (OSA) is vital for retail success, directly impacting profitability and customer satisfaction. However, traditional methods like manual stock checks and RFID sensors often lack accuracy and efficiency, leading to delayed restocking and poor inventory management. This research introduces an innovative semi-supervised learning framework integrated with the YOLO object detection model to automate empty-shelf detection. By minimizing manual annotation efforts, the system delivers robust performance across diverse retail environments. The optimized YOLOv5s variant achieves exceptional accuracy, boasting a mean Average Precision (mAP) of 99.2% at an Intersection over Union (IoU) threshold of 0.5, a precision of 95.6%, and a recall of 98.8%. Designed for real-time operation on edge devices, it provides a scalable, cost-effective solution for dynamic shelf management. Explainable AI (XAI) enhances transparency by offering visual insights into model decisions and detected shelves with empty spaces. The system also predicts restocking times, reducing manual effort and processing time. Validated across multiple datasets, this approach outperforms existing methods, setting a new benchmark for automated inventory management.
Keywords:
On-shelf availability
Deep learning
CNN
YOLO
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W