The Object Detection of Industry Nuts and Bolts Using Image Processing

Abstract

This paper presents an innovative deep learning approach for classifying nuts into two categories: good and broken. The model utilizes the MobileNetV2 architecture and is trained on a dataset consisting of annotated nut images. The achieved high accuracy underscores its efficacy for practical nut classification applications. The results contribute significantly to the progression of deep learning methodologies within the nut processing sector. The study presents an innovative deep learning-driven approach for image-based object detection, specifically emphasizing nut detection. The primary objective is to create a model proficient in accurately recognizing and localizing various nut types, such as hex nuts, within images. The project adopts the MobileNetV2 architecture as the foundational model, utilizing transfer learning with pre-trained weights from the Image Net dataset. The training dataset is composed of labeled nut images, annotated with bounding boxes. Training employs techniques like stochastic gradient descent and mini-batch optimization. Key metrics such as precision, recall, and F1-score are crucial for evaluating the model. The well-trained model can efficiently process new, unseen images in real-time or batch mode for nut detection. This project underscores the practical application of deep learning and objects detection in nut detection tasks, with potential implications in manufacturing, inventory management, and automated sorting processes.