ChartComplete: A Taxonomy-based Inclusive Chart Dataset

TL;DR

The ChartComplete dataset is presented, based on a chart taxonomy borrowed from the visualization community, and it covers thirty different chart types and does not include a learning signal.

Abstract

With advancements in deep learning (DL) and computer vision techniques, the field of chart understanding is evolving rapidly. In particular, multimodal large language models (MLLMs) are proving to be efficient and accurate in understanding charts. To accurately measure the performance of MLLMs, the research community has developed multiple datasets to serve as benchmarks. By examining these datasets, we found that they are all limited to a small set of chart types. To bridge this gap, we propose the ChartComplete dataset. The dataset is based on a chart taxonomy borrowed from the visualization community, and it covers thirty different chart types. The dataset is a collection of classified chart images and does not include a learning signal. We present the ChartComplete dataset as is to the community to build upon it.

Figures (4)

• Figure 1: The visual feature space of ChartComplete images. The features are extracted using Google VIT. The space is projected using T-SNE.
• Figure 2: The Centered Kernel Alignment (CKA) values between different chart types features. The features are extracted using Google ViT. The higher the CKA value the more the features are similar.
• Figure 3: The distribution of the image sizes in the ChartComplete dataset
• Figure 4: The distribution of collection methods per chart type.