STAIR Captions
收藏资源简介:
STAIR Captions是由千叶工业大学软件技术与人工智能研究实验室创建的大型日语图像描述数据集,包含820,310条日语描述对应164,062张图像。数据集基于MS-COCO图像,通过网络系统由约2,100名工作人员进行标注,遵循严格的标注指南。该数据集旨在支持日语图像描述的自动生成,解决现有数据集主要针对英语的问题,适用于图像搜索和视觉障碍人士的图像识别支持等领域。
STAIR Captions is a large-scale Japanese image captioning dataset developed by the Software Technology and Artificial Intelligence Research Laboratory of Chiba Institute of Technology. It contains 820,310 Japanese captions corresponding to 164,062 images. The dataset is built upon MS-COCO images, and was annotated by approximately 2,100 workers via an online system following strict annotation guidelines. This dataset aims to support the automatic generation of Japanese image captions, addressing the gap that most existing image captioning datasets primarily target English. It is applicable to fields such as image search and image recognition assistance for visually impaired people.




