Dataset Card for ImageNetVC

Dataset Summary

This is the dataset for our paper "ImageNetVC: Zero-Shot Visual Commonsense Evaluation on 1000 ImageNet Categories".

Supported Tasks

Zero- and Few-shot visual commonsense evaluation.

Languages

English

Dataset Structure

Data Instances

4076

Data Fields

Color, shape, material, component, and others.

Data Splits

Dev set for few-shot demonstrations and the main set for zero-shot evaluation.

Citation Information

If you find this dataset useful, please cite our paper:

@article{imagenetvc,
  author       = {Heming Xia and
                  Qingxiu Dong and
                  Lei Li and
                  Jingjing Xu and
                  Ziwei Qin and
                  Zhifang Sui},
  title        = {ImageNetVC: Zero-Shot Visual Commonsense Evaluation on 1000 ImageNet
                  Categories},
  journal      = {CoRR},
  volume       = {abs/2305.15028},
  year         = {2023},
  url          = {https://doi.org/10.48550/arXiv.2305.15028},
}