翻译：我们使用预训练的多模态预训练模型CLIP对实体名称与实体图像分别进行编码并计算相似度。对于实体图像我们直接利用CLIP进行图像编码而对于实体名称我们则将其修改为A photo of entity name再输入CLIP进行文本编码进而对相似度低于所设定阈值的图像进行删除。

We use the pre-trained multi-modal pre-training model CLIP to encode entity names and entity images separately and calculate their similarity. For entity images, we directly use CLIP for image encoding, while for entity names, we modify them to "A photo of entity name" and input them into CLIP for text encoding, and then delete images with similarity below the set threshold.