Extract raw data from plots images
Project description
A Python3 command line utility to digitize plots
This utility is useful when you have a lot of similar plots that needs to be digitized such as EEG, ECG recordings. See examples below.
Feel free to contact me for commercial work that may require optimizing this pipeline for your use case. Please send a sample plot.
For occasional use, have a look at WebPlotDigitizer by Ankit Rohatagi.
Installation
$ python3 -m pip install plotdigitizer
$ plotdigitizer --help
Preparing image
Crop the image and leave only axis and trajectories. I use
gthumb
utility on Linux. You can also use imagemagick or gimp.
Following image is from MacFadden and Koshland, PNAS 1990 after trimming. One can also remove top and right axis.
Run
plotdigitizer ./figures/trimmed.png -p 0,0 -p 10,0 -p 0,1
We need at least three points (-p
option) to map axes onto the image. In the example
above, these are 0,0
(where x-axis and y-axis intesect) , 20,0
(a point on
x-axis) and 0,1
(a point on y-axis). To map these points on the image, you
will be asked to click on these points on the image. Make sure to click in
the same order and click on the points as precisely as you could. Any error in
this step will propagate. If you don't have 0,0
in your image, you have to provide
4 points: 2 on x-axis and 2 on y-axis.
The data-points will be dumped to a csv file specified by --output /path/to/file.csv
.
If --plot output.png
is passed, a plot of the extracted data-points will be
saved to output.png
. This requires matplotlib
. Very useful when debugging/testing.
Notice the error near the right y-axis.
Using in batch mode
You can pass the coordinates of points in the image at the command prompt. This allows to run in the batch mode without any need for the user to click on the image.
plotdigitizer ./figures/trimmed.png -p 0,0 -p 20,0 -p 0,1 -l 22,26 -l 142,23 -l 22,106 --plot output.png
How to find coordinates of axes points
In the example above, point 0,0
is mapped to coordinate 22,26
assuming that
lower left corner is the origin of the image. I have included an utility
plotdigitizer-locate
(script plotdigitizer/locate.py
) which you can use to
find the coordinates of a data point.
$ plotdigitizer-locate figures/trimmed.png
or, from the source,
$ python3 plotdigitizer/locate.py figures/trimmed.png
This command opens the image in a separate window. You can click on a data point and its coordinate will be written on the image itself. Note them down.
Examples
plotdigitizer figures/graphs_1.png \
-p 1,0 -p 6,0 -p 0,3 \
-l 165,52 -l 599,51 -l 85,151 \
--plot figures/graphs_1.result.png \
--preprocess
plotdigitizer figures/ECGImage.png \
-p 1,0 -p 5,0 -p 0,1 \
-l 290,44 -l 1306,43 -l 106,301 \
--plot figures/ECGImage.result.png
Limitations
Currently this script has following limitations:
- One image should have only one trajectory.
- Grids are handled automatically. Every figure must have axes. Axes must not be tilted.
Need help
Open an issue and please attach the sample plot.
Related projects by others
- WebPlotDigitizer by Ankit Rohatagi is very versatile.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for plotdigitizer-0.2.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 7692d8c554cc1a20e67c4b1a2d9feb9851f5c68229dc905aca20dbdf6304dfed |
|
MD5 | 6a4a786b0a0c3808b00e0b20a7af037c |
|
BLAKE2b-256 | ab01882e69c8cf8db3fe0e343b6363244a42b0fd283321aeeacf788b49230856 |