Defects in composite manufacturing are common phenomena that can occur in multiple directions, serving as sources for crack propagation and significantly reducing material performance [1]. These defects pose a threat to structural integrity, potentially leading to premature material failure. Traditional methods for defect detection predominantly rely on image processing techniques [2], but variations in image quality often hinder their accuracy and reliability [3]. This study proposes a YOLOv8-based deep learning approach [4] to address the limitations of traditional image processing [2]. YOLOv8 achieves 95% accuracy in detecting, classifying, and segmenting defects, with detection in 81 ms and segmentation in 97 ms, even in low-quality images. This method enhances defect detection in industrial settings by identifying horizontal and vertical defects during fabrication. Real-time segmentation calculates defect areas, ensuring material integrity for optimal performance.