is there any clip model performing well on multiple objects

[yangzhj53]

I wanna use clip model to pick the right position relation of multiple objects or the right feature of a specific object in a muti-object picture, but most models seem to be weak to finish it. Is there any clip model trained on such special datasets or performing well in these situation?