Unet metrics

Continuing the discussion from Unet metrics:

Unfortunately, I could not follow this topic for a while and it was closed. But I have not received my answer yet.
This is my evaluation results:

"{'background': {'precision': 0.9443495, 'Recall': 0.9575637, 'F1 Score': 0.9509107377156112, 'iou': 0.90641546}, 'foreground': {'precision': 0.862103, 'Recall': 0.82460797, 'F1 Score': 0.8429387260488059, 'iou': 0.72851694}}"

On what threshold is this IoU calculated?
The output of the mask for each pixel is a number between zero and one, which for example considers the probability more than 0.5 to be 1 and the probability less than 0.5 to be 0.
I want to know this threshold.

There is no threshold. Use argmax to get the predicted value. And then compute the confusion_matrix based on groundtruth value and predicted value.

1 Like

Thank you @Morganh

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.