Image Processing OOP

Object-oriented digit image classification pipeline using TensorFlow/Keras.

This project includes data loading, preprocessing, model training, evaluation, prediction, and unit tests in a modular OOP structure.

Features

• OOP pipeline for loading and preprocessing grayscale digit images.
• CNN model builder with configurable activation, optimizer, and loss.
• Training and evaluation modules, including confusion matrix generation.
• Unit tests for core components.
• Notebook workspace for experiments.

Project Structure

.
|-- main.py
|-- requirements.txt
|-- README.md
|-- notebooks/
|   |-- Conv.Model.AI.ipynb
|   |-- Conv.Model.Finalipynb.ipynb
|   `-- test.ipynb
|-- extras/
|   |-- Images.csv
|   `-- OOP Dataloader.py
`-- image_processing/
	|-- __init__.py
	|-- data_object_final.py
	|-- dataloader.py
	|-- nn.py
	|-- train.py
	|-- evaluator.py
	|-- test.py
	|-- test_unit.py
	|-- digit_classifier.py
	|-- best_model.pt
	`-- Numbers_images_dataset/

Setup

1. Create and activate a Python virtual environment.
2. Install dependencies:

pip install -r requirements.txt

Run

Run the pipeline with default project-local paths:

python main.py

Useful options:

python main.py --skip-tests
python main.py --epochs 10
python main.py --dataset-zip image_processing/Numbers_images_dataset.zip
python main.py --dataset-dir image_processing/Numbers_images_dataset
python main.py --image-path image_processing/Numbers_images_dataset/3/0_0_7.jpeg

Tests

python -m unittest image_processing.test_unit

Notes

• The dataset and model artifacts can be large and are excluded through .gitignore.
• Notebooks are intended for experimentation; production logic is in image_processing/.