1 paper · 1 filter
Jiacheng Liu, Hang Zhou, Shida Wei +1
In this paper, we address the problem of plausible object placement for the challenging task of realistic image composition. We propose DiffPop, the first framework that utilizes p…