<![CDATA[On-shelf availability is a crucial aspect in the retail industry, directly impacting customer satisfaction and sales. Artificial intelligence-based object detection technology can enhance efficiency in monitoring product availability. This study examines the implementation of TensorFlow 2 for detecting retail products on shelves, using the SSD MobileNetV2 FPNLite architecture. Three model variations were developed based on input image sizes: 320x320, 640x640, and 1024x1024. The models were trained using transfer learning with a dataset containing 128 retail product classes. Evaluation results show that the 640x640 model achieved the best performance in terms of the trade-off between precision and speed, with a mAP of 0.72049 and an inference time of 0.283 seconds. The 320x320 model had the fastest inference time of 0.073 seconds, making it suitable for real-time applications. This study offers a solution to improve retail stock management through automatic object detection, aiming to reduce the risk of out-of-stock situations.]]>
Copyrights © 2024
