README.md

Recognize pattern in image with correlation

"One image can say more than a thousand word"

Features:

• Processing image by correlation (segmented by stage)
• Unitary test and Integration test
• Jupyter notebook for experiments and running without Editor/IDE
• Documented
• Clean code and simple design

Pre Requirements

• Python 3.6+
• PyCharm (recommended)

Running as command line

Open a terminal in the context of this repository, inside the virtual env, run the following command:

(venv)$ python3 main.py tests/img/all_numbers.png resources/img/numbers/ .png

You can also, export alias as alias recimg="python3 /home/user/dev/py-rec-img/main.py", and the use would be: (must be installed globally the requirements or use inside virtual env)

$ recimg tests/img/all_numbers.png resources/img/numbers/ .png

Steps:

• Open a terminal inside the virtual environment (venv)
• For install python required libs on env, run:

(env)$: pip install -r requirement.txt

Correlation Number Recognition Experiment

Recognize class pattern numbers in a image with samples of patterns.

Example target image:

Example class patterns:

The result should be the string with the number on the image (987654321).

-> Algorithm execution and explanation with Jupyter Notebook <-

Confusion Matriz of class patterns

How to

Open jupyter notebook

(env)$: pip install -r requirements-with-jupyter.txt
(env)$: jupyter notebook

Credits

Sham Vinicius Fiorin