A Yolov8-Nano Model for Reliable Real-Time Tyre Tread Defect Detection

This study proposes a lightweight deep learning framework based on YOLOv8-Nano for automated tyre tread defect detection, aiming to improve the reliability and efficiency of visual tyre condition monitoring. The proposed framework employs transfer learning with pre-trained COCO weights, a modified CSPDarknet backbone, and an anchor-free detection architecture to distinguish between normal and cracked tyre surfaces. Compared with conventional manual inspections, Tyre Pressure Monitoring Systems (TPMS), and computationally intensive deep learning approaches, the proposed method offers automated surface defect identification, reduced computational complexity, and potential suitability for real-time deployment in resource-constrained environments. To evaluate its effectiveness, a dataset comprising 400 labelled tyre images was subjected to preprocessing and data augmentation before being divided into training, validation, and testing subsets using a 70:15:15 ratio. The model was trained using the Adam optimiser and binary cross-entropy loss across four training configurations of 100, 120, 140, and 180 epochs. Performance was assessed through classification accuracy, precision, recall, F1-score, validation loss, and confusion matrix analysis. Experimental results demonstrated that the 180-epoch configuration achieved the most stable convergence, maintaining 100.00% Top-1 validation accuracy and a validation loss of 0.0000 over the final 70 epochs. Confusion matrix analysis further demonstrated correct classification of all 40 cracked and 40 normal tyre images, with no false positives or false negatives in the reported evaluation. These findings demonstrate the potential of YOLOv8-Nano as an efficient visual virtual sensor for automated tyre condition monitoring. Future integration with embedded automotive systems and temporal prediction models could support early defect identification, predictive maintenance, and intelligent transportation safety applications.