CNN Image Classification Repository

This repository contains the implementation of a Convolutional Neural Network (CNN) model for image classification. The project aims to achieve high accuracy in identifying classes of images through an efficient deep learning approach. The repository also incorporates a robust CI/CD pipeline for seamless integration, testing, and deployment.

Purpose

The CNN model is designed to classify images into predefined categories with high precision. We want to classify yes/no images that can be: circled, not circled, ticked and not ticked.

Model Details

• Architecture: The model is a custom-built CNN with the following layers:
  • Convolutional layers for feature extraction.
  • Max-pooling layers for dimensionality reduction.
  • Fully connected layers for classification.
  • Dropout layers to prevent overfitting.
• Activation Functions: ReLU for intermediate layers and Softmax for the output layer.
• Optimizer: Adam optimizer for efficient gradient descent.
• Loss Function: Cross-entropy loss to measure prediction error.
• Performance:
  • Training Accuracy:
  • Validation Accuracy:
  • Test Accuracy:

Achieving the Goal

1. Dataset:
   • The dataset consists of >1000 images categorized into 4 classes.
   • Preprocessing includes resizing, normalization. Data augmentation techniques such as flipping, rotation, and zooming has not been tried yet.
2. Training:
   • The model is trained on a GPU-enabled environment for faster computation.
   • Training involves multiple epochs with checkpoints for saving the best model.
3. Evaluation:
   • The model's performance is evaluated using accuracy, precision, recall, and F1-score.
   • Confusion matrix and classification report are generated for detailed insights.

Repository Structure

repo/
├── dataset/                 # folder with labelled images
├── .github/
│   ├── workflows               # CI/CD pipeline configuration
│   ├── release.yml             # workflow for automated semantic release tagging
│   ├── build_and_predict.yml   # workflow to train and run prediction on built model
├── results_model/
│   ├── folder with annotated images after prediction
├── README.md                # Project description and setup guide
├── requirements.txt         # Dependencies
├── .gitignore               # Ignored files and directories
├── main.py                  # utility functions
├── build_model.py           # script to build and save the model
├── run_predictions.py       # script to run prediction on a test dataset
├── notebook.ipynb           # Jupyter Notebook for testing and debugging

Stages of the Pipeline

1. Data Preprocessing:
   • Loads raw images and applies transformations such as normalization.
2. Model Training:
   • Defines and trains the CNN model.
   • Saves the best-performing model based on validation metrics.
3. Evaluation:
   • Evaluates the trained model on test data.
   • Generates performance metrics and visualizations such as confusion matrix.
4. Deployment:
   • Deploys the trained model using the CI/CD pipeline.
   • Monitors the deployment environment for performance issues.

How to Use

1. Clone this repository:

   git clone https://github.com/bruvio/image_detection.git
   cd repo

2. Install dependencies:

   pip install -r requirements.txt

3. Train the model:

   python build_model.py

4. Evaluate the model:

   python run_predictions.py

Contributions

Contributions are welcome! Please fork the repository and create a pull request with your proposed changes.

License

This project is open-source and available under the MIT License.

Author

• bruvio - Initial work - bruvio