README.md

---
language:
- en
license: apache-2.0
size_categories:
- 1K<n<10K
task_categories:
- visual-question-answering
pretty_name: ColorBench
dataset_info:
  features:
  - name: idx
    dtype: int64
  - name: id
    dtype: int64
  - name: type
    dtype: string
  - name: task
    dtype: string
  - name: filename
    dtype: string
  - name: image
    dtype: image
  - name: prompt
    dtype: string
  - name: question
    dtype: string
  - name: choices
    sequence: string
  - name: answer
    dtype: string
  - name: image_url
    dtype: string
  - name: image_type
    dtype: string
  splits:
  - name: test
    num_bytes: 3639277127.95
    num_examples: 5885
  download_size: 3535506580
  dataset_size: 3639277127.95
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
---

# 🎨 ColorBench
[**📖 Paper**](https://arxiv.org/abs/2504.10514) | [**💻 GitHub**](https://github.com/tianyi-lab/ColorBench)

ColorBench is a **multimodal dataset** to comprehensively assess capabilities of VLMs in color understanding, including color perception, reasoning, and robustness, introduced in ["ColorBench: Can VLMs See and Understand the Colorful World? A Comprehensive Benchmark for Color Perception, Reasoning, and Robustness"](https://arxiv.org/abs/2504.10514).

It provides:
- **More than 5,800 image-text questions** covering diverse application scenarios and practical challenges for VLMs evaluation.
- **3 categories and 11 tasks** for various color-centric capabilities evaluation including Perception (Color Recognition, Color Extraction and Object Recognition), Reasoning (Color Proportion, Color Comparison, Color Counting, and more) and Robustness. 

## 📃 Instruction
The data/test*.parquet files contain the dataset annotations and images pre-loaded for processing with HF Datasets.
```bash
from datasets import load_dataset

color_bench = load_dataset("umd-zhou-lab/ColorBench")
```

## 📂 Dataset Description

The dataset contains the following fields:

| Field Name | Description                                                                 |
|------------|-----------------------------------------------------------------------------|
| idx        | Global index of the sample in the dataset                                   |
| id         | Index of the sample in each task                                            |
| type       | Type of category: Perception, Reasoning, or Robustness                      |
| task       | Type of task: Color Recognition, Color Extraction, Color Counting, and more |
| filename   | Path to the image                                                           |
| image_url  | Source of the image                                                         |
| prompt     | Prompt with question and choices pre-formatted                              |
| question   | Question about the image                                                    |
| choices    | Answer choices for the question                                             |
| answer     | Correct answer to the question                                              |
| image      | Image object (PIL.Image)                                                    |