Visual Commonsense Graph (VCG)
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/Park-ing-lot/DIVE
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个大规模的视觉常识图，用于训练和评估模型在视觉常识生成任务上的表现。它包含了经过筛选的训练集以及几个验证子集，这些验证子集包括原始、独特和新颖的验证集。该数据集的任务是视觉常识生成。
This dataset is a large-scale visual commonsense graph designed for training and evaluating models' performance on visual commonsense generation tasks. It includes a filtered training set and several validation subsets, namely original, unique, and novel ones. The core task of this dataset is visual commonsense generation.
