Why do you ignore the void(label 11) instead of using it to classify the void pixels as "not relevant"?

alexgkendall / SegNet-Tutorial

Files for a tutorial to train SegNet for road scenes using the CamVid dataset

http://mi.eng.cam.ac.uk/projects/segnet/tutorial.html

847 stars 517 forks source link

Why do you ignore the void(label 11) instead of using it to classify the void pixels as "not relevant"? #100

Open mitalbert opened 7 years ago

mitalbert commented 7 years ago

I am planning to reduce the number of classes in cambid dataset and label the rest of the (not used)classes as void(say the same label 11). Now I wonder is it a good idea not to ignore this void pixels and instead train on this pixels to classify the "background"?

Would you please explain, why did you ignore the label 11 instead of using it?

ryanoa commented 6 years ago

@mitalbert Did you find a way to reduce the number of classes in camvid dataset easily? My team is trying to remove all the classes except 3. Is this something you achieved or similar?