No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
config Initial commit Jan 11, 2018
docs Initial commit Jan 11, 2018
figureqa import numpy as np for line 192 of (#1) Jul 11, 2018
resources Initial commit Jan 11, 2018
.gitignore Initial commit Jan 11, 2018
LICENSE.txt Initial commit Jan 11, 2018 Initial commit Jan 11, 2018
matplotlibrc Initial commit Jan 11, 2018
requirements.txt Initial commit Jan 11, 2018


Code to generate the FigureQA dataset, see

Data Generation

Data generation consists of 3 parts:

  1. Generate the source numerical data, styles, and question-answer pairs for the figures.
  2. Generate the figure images and bounding box annotatations.
  3. Aggregrate the figure images, questions & answers, annotations, and source data.

Code Map

All data generation source code lives in the figureqa/generation subpackage:

  • questions subpackage contains code to generate questions

    • for questions for bar graphs and pie charts.
    • for line plots.
    • for balancing and question encoding augmentation.
  • to generate source data, questions, and answers.

  • to generate figure images and bounding boxes.

  • aggregates the generated data into the documented format. Allows for generating a data split in multiple batches.

  • has misc. utilities for reconciling data formats, placing legends, etc.

  • defines the figure objects in Bokeh.

  • generates a whole dataset end-to-end.

  • generates images with bounding boxes visualized.

Each runnable module (script) can have its command line arguments displayed with --help.

There are some additional files used for data generation in these directories:

  • config contains .yaml files that configure visual apsects, source data parameters, color splits, and dataset generation.

  • resources contains the colors and other misc. resources for data generation.

And docs contains additional documentation on annotations, question format, and file formats.


  1. Install the FigureQA fork of Bokeh from
  2. pip install -r requirements.txt.
  3. Make sure you have enough space. The whole dataset unzipped is > 6GB, plus you need room for intermediate data.

Generate a whole dataset

Using a single script

This is done with the end-to-end script It does the source data synthesis, figure generation, and aggregation.

This script must be run from the root directory, FigureQA.

The config for the actual dataset is in config/figureqa_generation_config.yaml. A sample config is provided in config/sample_figureqa_generation_config.yaml.

Note that this does not generate the test sets.

With individual scripts

  1. cd FigureQA
  2. python figureqa/generation/ CONFIG_FILE.yaml SOURCE_DATA.json --<figure_type> <N_figures> ...
  3. python figureqa/generation/ SOURCE_DATA.json RAW_GENERATED_DIR
  4. python figureqa/generation/ FINAL_AGGREGATE_DIR RAW_GENERATED_DIR1 RAW_GENERATED_DIR2 ...