SegNet-Tutorial
SegNet-Tutorial copied to clipboard
What kind of work/complication would it be to extend with parly labeling
I am playing with an idea of having more training data, but only partly labeled, such each training image would have a unknown class label for those pixels that has not been labeled.
If anyone know or could estimate the workload for adding this to the tutorial, i would be happy.
I am working on a user experiences and i believe its easier to label more data with less labels, such humans can label the things they find easy and skip to the next image instead of forcing a label to each pixel. I dont have the level of expertise to see if this is easily made or not.