Recognition of Industrial Spare Parts Using an Optimized Convolutional Neural Network Model

Abstract

Spare parts search and retrieval processes are of paramount importance in manufacturing and supply chains. Image recognition using 2D and 3D image properties plays an important part in the success of such processes, as it facilitates the identification of the types and components associated with spare parts, a step that is crucial for their success. In this article, a novel Deep Learning-based object recognition model based on a convolutional neural network architecture is proposed and constructed using stacked convolutional layers to extract and learn features of the spare parts efficiently with the goal of improving the effectiveness of the spare part image recognition process. The proposed model is assessed using industrial spare parts datasets, and its performance is compared against different transfer learning models using precision, accuracy, recall, and F1 score. The proposed model demonstrated efficiency in spare parts recognition and achieved the highest accuracy compared to state-of-the-art image recognition models.