CLEVR-Humans
收藏arXiv2025-09-30 收录
下载链接：
https://cs.stanford.edu/people/jcjohns/iep/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个经过人工标注的集合，包含了关于各种物体图像的问题，这些物体以不同的形状、大小、材质和颜色呈现。此外，准确度是根据与地面真实答案的精确匹配来计算的。该数据集的任务是视觉问答。
This is a manually annotated dataset containing questions about images of various objects featuring diverse shapes, sizes, materials and colors. Additionally, accuracy is evaluated based on exact matching against ground-truth answers. The task of this dataset is visual question answering (VQA).
