Transformer-Based Multi-Objective Segmentation Network for Accurate Track Component Segmentation

To address these issues, this paper introduces a novel approach for segmenting track components at the pixel level using a multi-objective segmentation network based on Transformer. This network effectively segments multiple track components simultaneously, providing accurate pixel-level annotations for each component. The proposed method leverages the powerful capabilities of Transformer architecture to capture long-range dependencies and global context within the image, leading to improved segmentation performance.