--- dataset_info: features: - name: image dtype: image - name: id dtype: int64 - name: caption_ko dtype: string - name: captions_ko sequence: string - name: captions_en sequence: string splits: - name: train num_bytes: 13990978862.039 num_examples: 82783 - name: validation num_bytes: 6603067389.408 num_examples: 40504 download_size: 20150633676 dataset_size: 20594046251.447 configs: - config_name: default data_files: - split: train path: data/train-* - split: validation path: data/validation-* task_categories: - image-to-text - text-to-image language: - ko - en tags: - coco - image caption pretty_name: Ko COCO BAL size_categories: - 100K | | image | PIL을 이용해 읽어온 이미지 객체. 
| | captions_en | MS COCO의 원본 영문 설명 리스트. 
| | captions_ko | `captions_en`의 기계 번역본 리스트. 
| | caption_ko | `GPT-4o mini`를 이용해 `captions_ko`의 내용을 통합한 문장. 
|