Can Vision-Language Models Replace Human Annotators: A Case Study with CelebA Dataset [2410.09416]