Hi, thanks for the great work. I noticed from the paper that the labels assignment is set as follows:
Instead, we adopt a hard labelling strategy, whereby for a given pixel, we select the top class prediction of the teacher
network. We threshold the label based on teacher network output probability. Teacher predictions that exceed the
threshold become true labels, otherwise the pixel is labelled as ignore class. In practice we use a threshold of 0.9.
However I downloaded the Auto-Labelled data, I noticed that all pixels are annotated and there is no presence of ignore labels in this subset.
Hi, thanks for the great work. I noticed from the paper that the labels assignment is set as follows:
However I downloaded the Auto-Labelled data, I noticed that all pixels are annotated and there is no presence of ignore labels in this subset.