Enhancement of 3D Object Detection Using Deep Learning

Abstract

3D printing, an advanced form of additive manufacturing, has revolutionized production by enabling the creation of complex, customized objects used across industries like aerospace, healthcare, and automotive. Despite its benefits, 3D printing faces challenges such as defects (cracks, warping, surface imperfections) that compromise the structural integrity of printed objects, especially in high-precision applications. Traditional defect detection methods rely on manual inspection or image processing, which are time-consuming and error-prone. To address these issues, deep learning has been applied for automated defect detection. The proposed model, DenseNet201, is a pre-trained convolutional neural network (CNN), fine-tuned on a 3D printing defect dataset to classify defects and non-defects. Enhanced techniques, such as data augmentation, dropout regularization, and optimizer tuning (using the Adam optimizer), are implemented to optimize the model's performance. These methods contribute to the enhancement of 3D printing quality by improving defect detection accuracy. The trained model achieved maximum accuracy 93%, it also shows balanced performance across all classes with a score of 0.92 for average precision, recall, and F1 score; while showing the best performance on key defect types with F1 scores of 0.98 and 0.93 displaying strong defect detection which ultimately enhances manufacturing efficiency, reduces waste, and minimizes costs associated with traditional inspection methods. This approach aligns with the goal of utilizing deep learning to significantly improve the quality control process in 3D object detection.