README.md

metadata

task_categories:
  - image-text-to-text
license: apache-2.0
tags:
  - benchmark
  - multimodal
  - perception
  - computer-vision
dataset_info:
  features:
    - name: idx
      dtype: string
    - name: Question
      dtype: string
    - name: Image
      dtype: image
    - name: Groundtruth
      dtype: string
    - name: Task
      dtype: string
  splits:
    - name: full
      num_bytes: 171585064
      num_examples: 490
    - name: HiddenText
      num_bytes: 107144174
      num_examples: 150
    - name: 3DCaptcha
      num_bytes: 399420
      num_examples: 150
    - name: Colorblind
      num_bytes: 11984769
      num_examples: 150
    - name: ChineseLigature
      num_bytes: 52056701
      num_examples: 40
  download_size: 343068512
  dataset_size: 343170128
configs:
  - config_name: default
    data_files:
      - split: full
        path: data/full-*
      - split: HiddenText
        path: data/HiddenText-*
      - split: 3DCaptcha
        path: data/3DCaptcha-*
      - split: Colorblind
        path: data/Colorblind-*
      - split: ChineseLigature
        path: data/ChineseLigature-*

Turing Eye Test (TET) Dataset

Project Page: https://TuringEyeTest.github.io/ Code: https://github.com/TuringEyeTest/TuringEyeTest

Data Description

The Turing Eye Test (TET) as presented in Pixels, Patterns, but No Poetry: To See The World like Humans.

This benchmark includes four tasks:

• HiddenText: comprises scale-variant items where text is rendered as shapes within the figure, appearing as text when reduced and resolving into a complete image when magnified, which contains 150 images.
• 3DCaptcha: involves recognition challenges constructed with curved characters in the three-dimensional space, which consists of 150 Captchas.
• ColorBlind: similar to Ishihara tests, but augmented with confounding colored dots that are chromatically similar to the central character to increase difficulty. We generate 150 such test images.
• ChineseLigatures: features complex glyphs synthesized through character decomposition, morphological transformation, and fusion of multiple Chinese characters, which includes different words or phrases.

Citation

If you find this dataset is useful in your work, please cite our paper:

@misc{gao2025pixelspatternspoetryworld,
      title={Pixels, Patterns, but No Poetry: To See The World like Humans}, 
      author={Hongcheng Gao and Zihao Huang and Lin Xu and Jingyi Tang and Xinhao Li and Yue Liu and Haoyang Li and Taihang Hu and Minhua Lin and Xinlong Yang and Ge Wu and Balong Bi and Hongyu Chen and Wentao Zhang},
      year={2025},
      eprint={2507.16863},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2507.16863}
}