RTMOSeg: Robust Segmentation Method for Improved Accuracy

For example, in the segmentation results of U-net and DeepLabv3+, some parts of the background are classified as 'sleepers', while in the segmentation result of PSPnet, some parts of the background are identified as 'rails', leading to inadequate segmentation performance. However, our proposed segmentation method RTMOSeg can effectively identify each category and handle segmentation details accurately.