metadata

configs:
  - config_name: default
    data_files:
      - split: test
        path:
          - Objects.csv
          - Segmentations.json
          - visual_patterns.csv

PixMMVP Benchmark

The dataset annotations augmenting MMVP with referring expressions and corresponding segmentation masks for the objects of interest in their respective questions within the original VQA task.

Acknowledgements

I acknowledge the use of MMVP dataset's images and questions/choices part of building this dataset, the original MMVP.

References

Please city my work if you find the dataset useful

@article{siam2025pixfoundation,
  title={PixFoundation: Are We Heading in the Right Direction with Pixel-level Vision Foundation Models?},
  author={Siam, Mennatullah},
  journal={arXiv preprint arXiv:2502.04192},
  year={2025}
}