My-demo

U2-Net: U Square Net

• official repo of U2-Net: https://github.com/xuebinqin/U-2-Net
• In this demo U2-Net is used for portrait drawing

Virtual environment was created with the help of poetry:

• to install poetry according to your OS visit: https://python-poetry.org/docs/
• if you install poetry using pip, be aware that it will also install Poetry’s dependencies which might cause conflicts with other packages

Task 1

Follow these steps to run the project:

1. Go to My-demo directory
2. Activate virtual envirnonment with poetry run
3. Run python demo.py (can be different according to your OS)
4. You will see the result of portrait drawing based on the test photo
5. Then press exit to leave the shell

Task 2

First, if you don't have a build package, let's install it via command pythom -m pip install build

Second, to be able to build a package from the code:

1. poetry build
2. cd ./dist
3. pip install (pip install demo-0.1.0.tar.gz)
4. to watch demo you can run python demo.py

USE pip install git+https://github.com/whatswrongwithyourmitochondria/My-demo to install a package from the console

Task 3

• Styling tools such as black and pylint were added.

• pre-commit tool is used while making commits and has been configured to use styling tools mentioned above.

• to install all the necessary packages, run poetry install

Before commit

• to format the code with the help of black tool, run black demo.py
• to check the list of errors with the help of pylint tool, run pylint demo.py

to run all style checking tools at once:

• once run pre-commit install .
• every time you want to check your code with help of all styling tools, run pre-commit run -a

After you fix all the errors, make git add / git commit commands again.

Task 4

In this task 2 tests are added:

• visual regression test, that checks our results are always the same
• a "no error" test, that shows our algorithm works with the images of different sizes

To run tests, use the following command: pytest or python -m pytest

To check how CI works, follow the link below:

https://github.com/whatswrongwithyourmitochondria/My-demo/actions

CI installs dependencies, runs style checks (with the help of pre-commit that hooks black and pylint tools), tests code with pytest and then makes build and pushes it to the Artifacts.

You can also check artifacts selecting job in actions, e.g.:

https://github.com/whatswrongwithyourmitochondria/My-demo/runs/6675396697

You can scroll down and see Artifacts section. If needed, you can download a .zip archive:

Task 5

Web Demo for the project:

Web Demo is created with help of streamlit library. It is a single page application that takes image from a user and executes the demo code to draw the results in the browser.

You can find web_demo.py script in the /demo folder. In order to run this app, install the package (use local version of /dist folder) and then type streamlit run web_demo.py in /demo folder.

Streamlit app will work at localhost:8501.

You can choose .png image on your own and upload it.

Docker image

First, you need to install Docker to your system.

Second, run docker build -t unet ./

After building, use the following command

docker run -p 8501:8501 unet

and you'll see a web demo at localhost:8501

Take into account, that you may need to run those commands as administrator, otherwise access will be denied.