1 paper
Zitian Zhang, Frédéric Fortier-Chouinard, Mathieu Garon +2
We present ZeroComp, an effective zero-shot 3D object compositing approach that does not require paired composite-scene images during training. Our method leverages ControlNet to c…