sceneparsing
sceneparsing copied to clipboard
encoding for Red and Green channels
How exactly are the 150 object classes being encoded using the Red and Green channels?
We use only one channel to encode the 150 labels. For visualization, we use 3 channels.