SynthChartNet: 1.9M Synthetic Charts for Document AI

1,981,157 synthetically generated chart images with ground truth annotations form this multimodal dataset. Created by the docling-project and last updated in July 2025, it is designed for training the SmolDocling model on chart-based document understanding. Charts were rendered at 120 DPI using visualization libraries like Matplotlib, Seaborn, and Pyecharts.

Use Cases

• Train chart recognition models based on synthetically generated images of line, bar, and pie charts.
• Develop chart-to-text systems based on the OTSL-formatted ground truth annotations.
• Benchmark document AI models on a large-scale, diverse set of chart visualizations.
• Fine-tune vision-language models for tasks requiring comprehension of data visualizations.

Strengths

• Contains 1,981,157 samples, providing a large-scale resource for model training.
• Uses multiple visualization libraries (Matplotlib, Seaborn, Pyecharts), suggesting diversity in chart styles.
• Specifically designed for training the SmolDocling model, indicating a clear application focus.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the synthetic generation process and libraries used.

Provenance

Source

docling-project

Collection Method

Synthetically generated using visualization libraries.

Freshness

Last updated 2025-07-15 14:16:24