Code in this repo helps calculate a similarity score of two images.

The algorithm is based on MSE (Mean Squared Error) and a part of the Python code was obtained from this post.

TravisCI builds the python code into two executables; one for MacOS and another one for Windows; and uploads it to this Github repo under 'Releases' here.

• pytest unit tests are written in comparator_test.py. Running pytest from the command line will test basic functionality. It wasn't possible to integrate this into TravisCI because of an issue with opencv-python on Windows.
• You may also run python comparator.py to test functionality before committing changes. A sample images.csv and test images are provided images folder.

• A help dialog pops up every time the program is run.
• When the program finishes successfully, another dialog pops up letting the user know so.

• A new version is created everytime the build succeeds. A version info dialog pops up if a newer version of the program is available on Github. . Version info is bundled into the executable at build time by 'pyinstaller' and is checked on program startup.

• The #Maintainer section of this README contains setup and dev instructions for a future maintainer.

Download an executable for your operating system (MacOS or Windows) from the releases page.

In the same folder populate a CSV file with two columns:

• Column#1 must contain an absolute path to a source image.
• Column#2 must contain the absolute path to a second image to compare the first against.

Path must be compatible with the OS. Windows paths are of style: C:/Users/<-username->/Documents/pycompare/image1.png. MacOS paths are your typical Unix-style paths: /Users/<-username->Library/Documents/pycompare/image1.png

• The program doesn't yet accept files of GIF and TIFF formats due to opencv-python's limitations with licensing. This will be addressed in a future version.
• We're unable to run pytests because of a Windows-TravisCI bug. Hopefully, a fix is available in the future.
• A new version of the program is created on every successful TravisCI build, regardless of whether there's a change in the Python code. This must be fixed to only run on code changes.

Execute the program by double-clicking on the executable file. You will first be presented with a help dialog box. See: https://github.com/vsanjay85/pycompare#how-are-you-going-to-teach-bjorn-how-to-use-the-program

If the program runs successfully, you will be presented with a window citing so, and a file named processed_images.csv will be placed in the same folder where the executable was placed.

This file will contain four columns:

• Column1 and Column2 will be the same as the source CSV.
• Column3 holds the similarity score of the two source and target images. A score of '0' indicates 100% similar images.
• Column4 holds the time the program took to calculate the score, in seconds.

• Install Python 3.7 or up
• Open a terminal (Unix) or powershell (Windows) window
• Create a virtualenv, ensuring the venv is based on python 3.7
• Activate the virtualenv
• Clone this repo
• Install the required pip packages by running: pip3 install -r requirements.txt
• Create a new branch git checkout -b <branch-name>
• Make your modifications to the code
• Test the functioning of the code by running: python comparator.py. Code will use the sample images.csv and images from the images folder. Python works with both / and \ path separators without issues.
• If you'd prefer creating an executable to test, run the appropriate pyinstaller command from file .travis.yml. This creates an executable in the dist folder.
• Once tested and satisfied, push your changes to your branch git push.

A token is used to push the build artifacts into Github/Releases. This token is generated by using the travis CLI which is installed by running gem install travis Then you run: travis login --pro travis setup releases --com which automatically creates, encrypts, and adds the token to .travis.yml.