This figure is an extension of figure 2 from the paper. In this figure we show the results for the metrics: macro recall, macro precision, macro f1-score and accuracy. The premise of this figure is the same as described in the caption of figure 2 in paper.
This figure is an extension of figure 2 from the paper. In the paper, figure 2 only shows the macro recall for White Blood Cell dataset. However, In this figure we show the results for the Skin Lesion dataset and we show the following metrics: macro recall, macro precision, macro f1-score and accuracy. The premise of this figure is the same as described in the caption of figure 2 in paper.
This figure is an extension of figure 2 from the paper. In the paper, figure 2 only shows the macro recall for White Blood Cell dataset. However, In this figure we show the results for the Cell Cycle dataset and we show the following metrics: macro recall, macro precision, macro f1-score and accuracy. The premise of this figure is the same as described in the caption of figure 2 in paper.
This figure is an extension of figure 3 from the paper. In the paper, figure 3 shows that the combination of augmentation-based sampling, SimCLR or ImageNet pre-training and semi-supervised training with FixMatch is the optimal strategy on all three biomedical datasets. However, in the paper we only show macro recall. In this figure we show the results for the metrics: macro recall, macro precision, macro f1-score and accuracy.
This file contains the results of the complete grid search:
- For all three datasets: White Blood Cell, Skin Lesion, Cell Cycle
- All 5 active learning iterations
- Four metrics (mean and standard deviation): macro recall, macro precision, macro f1-score and accuracy
The results table shown in Table 2 in the paper.