Figure 3.

Confusion matrices

Confusion matrix comparing our classifier (left) versus the two different clinician types (middle and right) for classification of focused and panoramic images. Rows represent the correct label, while columns represent the label chosen by the classifier or the clinicians. The diagonal numbers represent the percent accuracy for each category (the percentage of time the correct label was identified), while the off-diagonal numbers represent misclassifications, with the number corresponding to the percentage of time the label for a given image type was ascribed to another, incorrect category.