Publication

Abstract : Managing electronic components efficiently in a laboratory environment poses a tedious problem, and the same can hinder research efficiency and productivity. These challenges stem from the diversity of electronic components, which can range from small passive components like resistors and capacitors to large integrated circuits. This research presents a groundbreaking system for managing electronic components in smart vending machines, leveraging advanced AI and innovative techniques. The Multimodal Annotation Fusion (MAF) method enhances a carefully chosen dataset with various attributes and multimodal annotations, laying the groundwork for a robust and intelligent recognition system. YOLO v8 and transfer learning are integrated along with custom loss functions and model optimisation to produce a model that significantly outperforms existing solutions. The algorithm’s effectiveness is particularly noticeable in its low GPU time consumption, which is important for real-time applications. Ablation experiments in various settings further validate the algorithm’s efficacy, particularly in identifying small electrical components. As an example, the suggested model shows a 3% higher IoU than SSD and a 1% higher IoU than YOLO v7 and Faster RCNN, indicating significant increases in accuracy. The results of this study could greatly improve electronic component management, increasing the usefulness of smart vending machines for professionals and students working in lab environments. Furthermore, the vending machine’s digital twin accurately simulates its real-world performance for virtual testing and improvement prior to real deployment, offering increased effectiveness and convenience.