Point-Cloud Completion with Pretrained Text-to-image Diffusion Models Y oni Kasten