Improving Biomedical Image Segmentation: An Extensive Analysis of U-Net for Enhanced Performance

The U-Net architecture is the focus of this study, which optimizes biomedical picture segmentation. Improving performance in contexts with limited resources is the goal. The methodology uses GradCAM++, k-fold cross-validation, and kernel size modifications to conduct a systematic evaluation over many folds. Significant gains in accuracy, precision, recall, F1- score and accuracy are indicate in the results which reach 96.70%. By lowering the reliance on data augmentation, the suggested modifications improve the model's effectiveness. Under resource constraints, our work provides important new under- standings for biological picture segmentation.