Real-time warp and weft angle detection in fabrics using an oriented bounding box–based YOLOv8 deep learning model

Finishing processes such as dyeing, washing, drying, and shaping applied during fabric production may distort the textile structure, leading to angular deviations between warp and weft yarns. Weft straightening machines (WSMs) are employed in industrial production lines to monitor and correct these deviations; however, existing systems often exhibit limited robustness and accuracy under high-speed operating conditions. Undetected small angular errors may therefore evolve into irreversible structural defects, particularly during drying and finishing stages. In this study, a deep learning–based weft skew detection approach is proposed for real-time angle estimation in industrial WSMs using an oriented bounding box (OBB)-based object detection framework. The YOLOv8-OBB architecture is adopted due to its ability to jointly estimate object localization and rotation with high computational efficiency, making it suitable for real-time industrial applications. The model is trained and evaluated on images of five fabric types processed at production speeds ranging from 10 to 130 m/min. Experimental results demonstrate that the proposed method achieves mAP@0.5 values between 85% and 91%, enabling accurate and reliable detection of warp and weft orientations. The findings indicate that the proposed approach provides an effective and scalable solution for real-time quality control and angle-based correction in textile finishing lines.