Railway Track Anomaly Detection Based on EfficientNet-B0 and SimAM

To address the need for model lightweighting and improved classification performance in railway track anomaly detection, this study proposes an image classification algorithm based on EfficientNet-B0 and the SimAM attention mechanism. By introducing the lightweight and efficient EfficientNet-B0 as the backbone model and integrating the SimAM attention mechanism, the proposed method enhances the model's focus on target regions while suppressing irrelevant background information, thereby improving classification accuracy. EfficientNet-B0 employs MBConv modules and a compound scaling strategy, enabling significant reductions in model parameters and computational complexity while maintaining high classification performance. Experimental results demonstrate that the proposed method outperforms other mainstream models on multiple public datasets, validating its efficiency and robustness in image classification tasks.