Accurate online inspection of train wheelset tread defects is challenging owing to the variety and position uncertainty of defects. This study develops an improved YOLOv7 model capable of inspecting various wheelset tread defects with high accuracy and low computation complexity. This model comprises GSConv, a small target enhancement (STE) module, and StyleGAN3. GSConv significantly reduces the model volume while maintaining the feature expression ability, achieving a lightweight structure. The STE module enhances the fusion of shallow features and distribution of attention weights, significantly improving the sensitivity to small-sized defects and positioning robustness. StyleGAN3 enhances small samples by addressing inhomogeneity, thereby generating high-quality defect samples; it overcomes the limitations of traditional amplification methods regarding texture authenticity and morphological diversity, systematically improving the model’s generalization ability under sample scarcity conditions. The model achieves 1.6%, 10.7%, 48.63% and 37.97% higher mean average precision values than YOLOv7, YOLOv5, SSD, and Faster R-CNN, respectively, and the model parameter size is reduced by 73.91, 94.69, 122.11, and 154.91 MB, respectively. Hence, the proposed YOLOv7-STE model outperforms traditional models. Moreover, it demonstrates satisfactory performance in detecting small target defects in different samples, highlighting its potential applicability in online wheel tread defect inspection.

Train wheelsets are key components of high-speed trains and are prone to failure during long-term operation under harsh conditions. Accurate and comprehensive online inspection of defects in wheelset treads are difficult because of their variety and position uncertainty. Accurate detection and location of small targets are highly challenging. This study developed an improved YOLOv7 model with better accuracy and lower power computation. The model was applied to the inspection of various wheelset-tread defects. GSConv was used to reduce the model volume, while a small target enhancement module addressed low pixels and difficulty in distinguishing small targets. This improved the classification accuracy for damage feature recognition. The captured images of the wheelset tread were preprocessed using StyleGAN3 to augment small samples and address non-uniformities. The experimental results showed that the proposed model exhibited a remarkably better overall performance than conventional models, particularly in terms of mean average precision. The public RSDDs dataset of small rail defects was selected to verify the robustness of the proposed algorithm. Compared with the traditional network model, the improved model proposed herein showed good performance for small target defects in different samples, highlighting its potential applicability in online inspection of wheel tread defects.