Researchers are exploring various approaches to improve segmentation tasks, including boundary-aware polyp segmentation and skin lesion segmentation. The SegFormer model is being investigated for its potential in these areas, with some studies focusing on its application in medical imaging and others in metal corrosion detection. Different methods, such as adaptive multi-branch fusion and multi-modal attention networks, are being developed and tested to enhance the model's performance.