Fig. 4

Visualization of U-Net segmentation predictions. The columns display (from left to right) the following: original dermatoscopic images, ground truth masks, and predicted masks. The results highlight the model’s ability to segment skin lesions with complex boundaries and varying textures accurately.