"ChartComplete: A Taxonomy-based Inclusive Chart Dataset"

"ChartComplete is a taxonomy-based inclusive chart image dataset designed to support research in chart classification, chart understanding, and multimodal learning. The dataset is constructed using a modified version of Borkin\u2019s visualization taxonomy and covers 30 distinct chart types across 12 categories, including common charts (bar, line, pie, scatter) and specialized visualizations (parallel coordinates, box and whisker plots, Sankey diagrams, choropleth maps, radar charts, heat maps, and tree maps).The dataset contains 1,500 high-quality chart images, with 50 images per chart type. Images were collected through a hybrid approach combining automated web scraping (Statista and Our World in Data) and manual curation. All images satisfy strict quality criteria, including minimum resolution requirements, readable text, complete labeling (titles, axes, legends), and real-world data authenticity.ChartComplete is organized hierarchically by category and chart type to enable systematic evaluation of machine learning models across visualization families. The dataset supports research in:Chart type classificationVision-language model evaluationChart question answering (ChartQA)Transfer learning for visual understandingRepresentation learning for visualizationThe dataset is distributed under the Creative Commons Attribution 4.0 (CC BY 4.0) license."