mardausdennis/image-classification-project

A machine learning project for classifying images into 20 categories using a custom CNN model implemented in PyTorch.

Image Classification Project

This project is for the Programming in Python II course, focusing on building a machine learning model to classify images into 20 different categories.

Project Structure

• data/: Contains the dataset files (images and labels).
  • training_data/: Directory where the training images and CSV file are stored.
  • validation_indices.npy: Numpy file storing the indices for the validation set.
• models/: Contains the saved model files.
• src/: Contains the source code files.
  • architecture.py: Contains the model architecture.
  • dataset.py: Contains data loading and preprocessing code.
  • train.py: Contains code for training the model.
  • evaluate.py: Contains code for evaluating the model.
• requirements.txt: Lists the dependencies required for the project.
• README.md: Project overview and setup instructions.

Setup Instructions

1. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

2. Install dependencies:

   pip install -r requirements.txt

3. Download and place the dataset in the data/training_data/ directory.

Usage

• To train the model, run:

  python src/train.py

• To evaluate the model, run:

  python src/evaluate.py

Notes

• The dataset is expected to be in grayscale, and the model architecture and training process have been set up accordingly.
• The validation_indices.npy file is used to separate the training and validation datasets.