1 paper
Yilun Du, Kevin Smith, Tomer Ulman +2
We study the problem of unsupervised physical object discovery. While existing frameworks aim to decompose scenes into 2D segments based off each object's appearance, we explore ho…