Electronic components recognition based on deep learning

Abstract

Deep learning methods, particularly Convolutional Neural Networks (CNNs), have revolutionized image classification, including the classification of electronic components, when compared to traditional methods, deep learning methods have several advantages. Firstly, CNNs are capable of automatically learning hierarchical representations from raw input data, eliminating the need for manual feature engineering. This allows CNNs to effectively capture intricate patterns and features in images, enabling accurate classification. Moreover, it excels in handling large-scale datasets. In this thesis, we focused on investigating the classification of electronic components using deep learning techniques and evaluated the performance of popular models such as LeNet, AlexNet, GoogleNet, and VGG16, leveraging transfer learning by comparing their performance on a small dataset representing four classes of electronic components: bypass capacitor, transistor, LED, and relay. Notably, our findings demonstrated that VGG-16 achieved superior results, exhibiting higher accuracy within a shorter time frame. This outcome highlights the effectiveness of transfer learning, where pre-trained models can be fine-tuned on specific tasks, in improving the classification accuracy of electronic components.