Generates synthetic datasets for training and evaluating vision models on chart analysis and extreme value identification tasks. Each sample contains a chart without explicit data labels where the extreme point (maximum or minimum) must be identified visually and highlighted.

Each sample pairs a task (first frame + prompt describing what needs to happen) with its ground truth solution (final frame showing the result + video demonstrating how to achieve it). This structure enables both model evaluation and training.

# 1. Clone the repository
git clone https://github.com/VBVR-DataFactory/G-30_chart_extreme_without_data_data-generator.git
cd G-30_chart_extreme_without_data_data-generator

# 2. Create and activate virtual environment
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# 3. Install dependencies
pip install --upgrade pip
pip install -r requirements.txt
pip install -e .

# Generate 50 samples
python examples/generate.py --num-samples 50

# Custom output directory
python examples/generate.py --num-samples 100 --output data/my_dataset

# Reproducible generation with seed
python examples/generate.py --num-samples 50 --seed 42

# Without videos (faster)
python examples/generate.py --num-samples 50 --no-videos

The scene shows a area chart. Find the minimum value point and draw a red rectangular border around the corresponding point to highlight it.

Initial Frame Bar chart without data labels	Animation Red border appears around longest bar	Final Frame Maximum value bar highlighted

Analyze charts without explicit data labels to identify extreme values (maximum or minimum) visually by comparing bar lengths or point positions, then highlight the target.

• Chart types: Bar charts (horizontal/vertical), line charts, pie charts, scatter plots, area charts
• Data visibility: No explicit numerical labels on data points
• Visual comparison: Must compare bar lengths or point heights visually
• Axes: Basic axes without detailed value labels
• Color palette: 30 distinct colors for chart elements
• Extreme type: Maximum (longest/highest) or minimum (shortest/lowest)
• Highlight method: Red rectangular border around target bar/point
• Background: White with minimal chart elements
• Goal: Identify correct extreme element purely by visual comparison

• Visual-only data comparison (no numerical labels)
• Extreme value identification through spatial reasoning
• Multiple chart type support (horizontal bar, vertical bar, line, pie, scatter, area)
• 30 distinct colors for enhanced visual diversity
• Tests visual comparison and spatial reasoning skills
• More challenging than with-data version
• Red border annotation for highlighting
• Complete metadata with per-point color and value information

data/questions/chart_extreme_without_data_task/chart_extreme_without_data_00000000/
├── first_frame.png      # Chart without data labels or highlighting
├── final_frame.png      # Chart with extreme element highlighted
├── prompt.txt           # Instruction to find and highlight extreme value
├── ground_truth.mp4     # Animation of highlight appearing
└── question_metadata.json # Task parameters and metadata

File specifications:

• Images: 1024×1024 PNG format
• Video: MP4 format, 16 fps
• Duration: ~3 seconds

Metadata structure: Each sample includes question_metadata.json with the following parameters:

• chart_type: Type of chart (e.g., "horizontal_bar", "bar", "line", "pie", "scatter", "area")
• values: List of data values
• data_points: List of objects, each containing:
  • index: Data point index (0-based)
  • value: Numerical value
  • color: RGB color tuple [r, g, b] assigned to this data point
• extreme_type: "max" or "min"
• target_index: Index of the extreme value
• target_value: The extreme value itself

logic-symbols chart-analysis visual-comparison extreme-values spatial-reasoning visual-highlighting