We applied a deep learning technique to detect defects on the glass used in a thin-film transistor liquid crystal display (TFT-LCD) utilizing a You Only Look Once v4 (YOLOv4) object detection model. TFT-LCD glass defect detection is a critical quality control step in electronics manufacturing. Defects on the glass indicate serious problems in production. Manual inspections are often inefficient and inconsistent, highlighting the need for automated methods. To enhance efficiency and accuracy in the automated detection of defects on the TFT-LCD glass, convolutional neural networks (CNNs) were used. By optimizing and training the YOLOv4 model with a large labeled dataset, a highly efficient object detection method for multiple defects was developed. CNNs based on YOLOv4 showed superior performance in real-time detection and reduced defect detection time. Additionally, smart sensor CCD technology was employed to capture high-resolution images of glass surfaces for precise defect detection. The model leverages deep learning concepts such as feature extraction, data augmentation, and loss function optimization to improve performance. The developed YOLOv4 object detection model can be used for the quality control of automated TFT-LCD production and can help increase production efficiency and reduce defects of the final products.