Figure 4
From: Impact of image compression on deep learning-based mammogram classification
Performance of models trained with compression-based data augmentation. The plot shows AUROC scores of models trained using different CR ranges. Each label in the legend indicates a model trained on data with a mixture of CRs. For example, M-CR1toCR11K refers to the model trained on a dataset with a mixture images with different CRs ranging between 1 and 11 K. Each model is evaluated on images with a single CR of 1, 15, 20, 25, 50, 100, 500, 1 K, 5 K, 10 K, and 11 K, respectively.
