MoEdit: On Learning Quantity Perception for Multi-object Image Editing | Read Paper on Bytez