googlyeyes

A lightweight web application which adds googly eyes to an image sent to its POST /imageUpload endpoint.

`POST` payload	Response body	Faces
		`n_faces=1`
		`n_faces=2`
		`n_faces=3`

Installation & Usage

Install in editable mode using pip:

cd googlyeyes/
conda create -n googlyeyes-env -c conda-forge python=3.6 coverage flake8 Flask Flask-RESTful imutils dlib opencv numpy pytest requests
conda activate googlyeyes-env

Building the docker image

docker build -t googlyeyes-app .

Run unit and regression tests

You can then run the application using:

docker run -d -p 5000:5000 googlyeyes-app
pytest

Code coverage

Name                             Stmts   Miss  Cover   Missing
--------------------------------------------------------------
googlyeyes/helper_functions.py      23      0   100%
googlyeyes/process_image.py         48     14    71%   149-178, 182
googlyeyes/server.py                34     15    56%   21-22, 26-38, 47-53, 64-65
tests/test_helper_functions.py      36      0   100%
tests/test_process_image.py         30      0   100%
tests/test_server.py                47      0   100%
--------------------------------------------------------------
TOTAL                              218     29    87%

Performance testing

Synchronous performance tests

Performance tests were made by making 1000 sequential requests to a single docker image running the application:

Statistic	Response time (ms)
Mean	354.90
99.9%	829.52
99.999%	845.59

Asynchronous performance tests

To-do for v0.4.

Facial coordinate points

68 facial coordinate points from the iBUG 300-W dataset.

Credits

Credits to pyimagesearch for the code to find the facial landmarks.
Credits to sagaragarwal94 for his simple webserver API using Flask.
Credits to fireant for his StackOverflow answer on overlaying a PNG image with OpenCV.
Credits to the Helen dataset which was used for testing.

manisoftwartist/googlyeyes