Training-free Cross-domain Few-shot Segmentation via Robust Semantic Representation and Matching | ResearchPod