1 paper · 1 filter
Shreyas Chandgothia, Ardhendu Sekhar, Amit Sethi
Training a computer vision system to segment a novel class typically requires collecting and painstakingly annotating lots of images with objects from that class. Few-shot segmenta…