README.md

metadata

license: cc-by-nc-4.0
language:
  - ro
dataset_info:
  features:
    - name: id
      dtype: string
    - name: images
      sequence: image
    - name: label
      dtype: string
    - name: count
      dtype: int64
    - name: points
      struct:
        - name: x
          sequence: float64
        - name: 'y'
          sequence: float64
    - name: messages
      list:
        - name: role
          dtype: string
        - name: content
          dtype: string
  splits:
    - name: train
      num_bytes: 6200055538.732
      num_examples: 34579
  download_size: 6306784557
  dataset_size: 6200055538.732
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Dataset Description

PixmoCount is a dataset of images paired with number of objects in the image.

Here we provide the Romanian translation of the PixmoCount dataset, translated with Seed-X-PPO. This dataset is part of the instruction finetune protocol for Romanian VLMs proposed in "Înțelegi românește?" A Recipe for Romanian Vision-Language Models (Masala et al., 2026).

Citation

@inproceedings{deitke2025molmo,
  title={Molmo and pixmo: Open weights and open data for state-of-the-art vision-language models},
  author={Deitke, Matt and Clark, Christopher and Lee, Sangho and Tripathi, Rohun and Yang, Yue and Park, Jae Sung and Salehi, Mohammadreza and Muennighoff, Niklas and Lo, Kyle and Soldaini, Luca and others},
  booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
  pages={91--104},
  year={2025}
}

@misc{masala2026intelegi,
      title={``\^{I}n\c{t}elegi Rom\^{a}ne\c{s}te?'' A Recipe for Romanian Vision-Language Models},
      author={Mihai Masala and Marius Leordeanu and Mihai Dascalu and Traian Rebedea},
      year={2026},
      eprint={2605.31401},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2605.31401},
}