Dear PraNet Team,
I use your network to make segmentations on my images. Some images contain a ground truth, that can be segmented, some images are true negatives and the model should not make a prediction (mask should be all black). When I have the true positives the mask is very good on the test prediction. When I have a true negative the model does this:
Is this a prediction problem or a postprocessing problem with the image saving and the sigmoid/min max function?
Thank you for your help and best regards
Dear PraNet Team,
I use your network to make segmentations on my images. Some images contain a ground truth, that can be segmented, some images are true negatives and the model should not make a prediction (mask should be all black). When I have the true positives the mask is very good on the test prediction. When I have a true negative the model does this:
Is this a prediction problem or a postprocessing problem with the image saving and the sigmoid/min max function?
Thank you for your help and best regards