piechartocr 0.6.6

Pie Chart Optical Character Recognition

pie-chart-ocr

A tool to extract tabular data from pie charts, developed as a component of the CryptoSearchTools toolkit.

Note: The original repository was moved to https://git.ehtec.co/research/pie-chart-ocr. https://github.com/ehtec/pie-chart-ocr is a mirror.

Installation

Install via PyPi

You can install all tagged versions of piechartocr from PyPi:

python3 -m pip install --upgrade piechartocr

Note: You cannot run tests and examples from the PyPi installation. The required files need to be downloaded from Gitlab.

Build from source

Install Boost and Tesseract:

sudo apt install libboost-system-dev tesseract-ocr build-essential git

Clone this repository including submodules:

git clone --recursive https://github.com/ehtec/pie-chart-ocr.git
cd pie-chart-ocr

Install Python requirements:

python3 -m pip install -r requirements.txt

Compile libraries:

python3 setup.py build_ext

Create temporary directories:

mkdir temp
mkdir temp1
mkdir temp2

Unpack test charts:

unzip data/charts_steph.zip -d data
unzip data/charts_steph_upsampled.zip -d data
unzip data/generated_pie_charts_legend.zip -d data
unzip data/generated_pie_charts_without_legend.zip -d data

Usage

Run unit tests:

python3 -m nose2 --start-dir tests/ --with-coverage

Run legacy tests / examples:

python3 run_examples.py

Generate test data (mock pie charts):

python3 run_generate_test_data.py

To extract data from any pie chart:

from piechartocr import pie_chart_ocr

# Path to pie chart
path = "/path/to/my/chart.png"

# Extract data
data = pie_chart_ocr.main(path, interactive=False)

# Print the extracted list of tuples of the form [(percentage / 100, label)]
print(data["res"])

Metrics

These metrics are autogenerated by the CI-pipeline.

Metrics for mock pie charts with legend:

Metrics for mock pie charts without legend:

Metrics for real world pie charts (many of them in awful quality, some even unreadable for humans):