CLoDSA: a tool for augmentation in classification, localization, detection, semantic segmentation and instance segmentation tasks

Table 4 Results using several models trained with different datasets (D₁ is the original dataset, D₂ is D₁ augmented using elastic deformations, D₃ is D₁ augmented using geometric and color transformations, and D₄ is the combination of D₂ and D₃) for the ISBI challenge

	Accuracy	F1-score	Precision	Recall	Specificity	Balanced accuracy
D ₁	0.90	0.94	0.93	0.94	0.76	0.85
D ₂	0.90	0.94	0.94	0.94	0.77	0.855
D ₃	0.90	0.94	0.95	0.92	0.82	0.87
D ₄	0.91	0.94	0.94	0.94	0.78	0.86

ISSN: 1471-2105