1 paper
Elias Kassapis, Georgi Dikov, Deepak K. Gupta +1
In semantic segmentation tasks, input images can often have more than one plausible interpretation, thus allowing for multiple valid labels. To capture such ambiguities, recent wor…