Figure 11.

A segmentation example from INbreast by Deeplabv3+ with different backbones. (a) Pre-processed image. (b) $D{L}_{ResNet18}$: 97.99% of IoU, 99.80% of GPA, 96.85% of DSC, 93.89% of Jaccard, 100% of Sensitivity, and 99.60% of Specificity. (c) $D{L}_{ResNet50}$: 98.60% of IoU, 99.86% of GPA, 97.39% of DSC, 94.92% of Jaccard, 100% of Sensitivity, and 99.72% of Specificity. (d) $D{L}_{MobileNetv2}$: 98.15% of IoU, 99.82% of GPA, 96.82% of DSC, 93.83% of Jaccard, 100% of Sensitivity, and 99.63% of Specificity. (e) $D{L}_{XceptionNet}$: 98.87% of IoU, 99.83% of GPA, 98.08% of DSC, 96.23% of Jaccard, 98.04% of Sensitivity, and 99.99% of Specificity. (f) $D{L}_{InceptionResNetv2}$: 99.21% of IoU, 99.88% of GPA, 98.28% of DSC, 96.62% of Jaccard, 99.89% of Sensitivity, and 99.86% of Specificity.