RTMOSeg: Robust Object Segmentation Model for Railway Infrastructure

In the same category, the RTMOSeg model has the highest test accuracy. These results show that the proposed RTMOSeg model has a better segmentation effect than other models. Additionally, for different types, the test accuracy of fasteners and sleepers in other models is much lower than that of rails, while the test accuracy of RTMOSeg is relatively small. The experimental results demonstrate that RTMOSeg is effective for segmenting both simple and complex targets. Table 3 displays the test results of each model trained on dataset 2. We noticed that the test accuracy of RTMOSeg remains the highest within the same category. In multiple comparison methods, the test accuracy of fasteners and sleepers in multiple models is much lower compared to that of dataset 1, and the test accuracy of sleepers is lower than that of fasteners. This is because the shape and position of fasteners and sleepers become more complex in images containing multiple rails (i.e., images with turnouts). When compared to other models, the RTMOSeg model proposed in this paper achieves the highest test accuracy, and there is little difference in the test accuracy between dataset #1 and dataset #2, further proving the strong robustness of RTMOSeg.