Datasets:

Modalities:
Image
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
pandas
License:
joliang17 commited on
Commit
a9dd732
Β·
verified Β·
1 Parent(s): 89ad260

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +8 -7
README.md CHANGED
@@ -12,8 +12,8 @@ dataset_info:
12
  dtype: string
13
  - name: filename
14
  dtype: string
15
- - name: image_url
16
- dtype: string
17
  - name: prompt
18
  dtype: string
19
  - name: question
@@ -22,8 +22,8 @@ dataset_info:
22
  sequence: string
23
  - name: answer
24
  dtype: string
25
- - name: image
26
- dtype: image
27
  splits:
28
  - name: test
29
  num_bytes: 1808612125.49
@@ -45,12 +45,13 @@ size_categories:
45
  ---
46
 
47
  # 🎨 ColorBench
 
48
 
49
- 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".
50
 
51
  It provides:
52
- - **5,800 image-text questions** covering various image types and applications for VLMs evaluation.
53
- - **3 categories and 11 tasks** for multiple color-related capabilities evaluation including Perception (Color Recognition, Color Extraction and Object Recognition), Reasoning (Color Proportion, Color Comparison, Color Counting, and more.) and Robustness.
54
 
55
  ## πŸ“ƒ Instruction
56
  The data/test*.parquet files contain the dataset annotations and images pre-loaded for processing with HF Datasets.
 
12
  dtype: string
13
  - name: filename
14
  dtype: string
15
+ - name: image
16
+ dtype: image
17
  - name: prompt
18
  dtype: string
19
  - name: question
 
22
  sequence: string
23
  - name: answer
24
  dtype: string
25
+ - name: image_url
26
+ dtype: string
27
  splits:
28
  - name: test
29
  num_bytes: 1808612125.49
 
45
  ---
46
 
47
  # 🎨 ColorBench
48
+ [**πŸ“– Paper**](todo) | [**πŸ’» GitHub**](https://github.com/tianyi-lab/ColorBench)
49
 
50
+ 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"](todo).
51
 
52
  It provides:
53
+ - **More than 5,800 image-text questions** covering diverse application scenarios and practical challenges for VLMs evaluation.
54
+ - **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.
55
 
56
  ## πŸ“ƒ Instruction
57
  The data/test*.parquet files contain the dataset annotations and images pre-loaded for processing with HF Datasets.