Binary Galaxy Classification with EfficientNet-B0 and PyTorch

This project implements a binary classification model to distinguish between spiral and elliptical galaxies using the EfficientNet-B0 architecture and the PyTorch framework. The model was trained on a subset of the Galaxy Zoo dataset and achieved high accuracy in classifying galaxy images.

Project Structure:

├── utils
│   ├── prediction.py
│   ├── model.py
│   └── transformations.py
└── app.py

File Description:

• app.py: Contains the main code for the Gradio interface, which allows users to upload images and obtain predictions from the model.
• utils/model.py: Defines the EffNet class, which loads a pre-trained EfficientNet-B0 model and adds a dropout layer and a fully connected layer for binary classification.
• utils/prediction.py: Includes the predict_class function that receives an image tensor and returns the predicted class (elliptical or spiral) and the confidence of the prediction.
• utils/transformations.py: Defines the transform_image function, which loads an image from the specified path, applies the necessary transformations and returns the image tensor ready for use by the model.

Dependencies:

The project requires the following Python libraries:

• torch
• torchvision
• efficientnet_pytorch
• Pillow
• gradio
• scikit-learn

You can install the dependencies using pip:

pip install -r requirements.txt

How to Run the Project:

1. Clone the repository:

git clone https://github.com/Rajora0/binary-classification-galaxies.git

2. Navigate to the project directory:

cd binary-classification-galaxies

3. Run the Gradio application:

python app.py

4. Access the Gradio interface in your browser, at the address provided in the terminal.

5. Upload an image of a galaxy and click "Submit." The model will classify the image as elliptical or spiral and display the predicted class and the confidence of the prediction.

Training a Custom Model:

The project includes Jupyter notebooks that guide you through the process of training your own custom model. Here's how:

1. Data Preparation:

   • Download the Galaxy Zoo dataset or use your own dataset of galaxy images.
   • Organize your data into separate folders for each class (e.g., "spiral" and "elliptical").
   • Use the provided data loading and preprocessing functions to prepare your data for training.

2. Model Training:

   • Open the training.ipynb Jupyter notebook.
   • Modify the notebook to load your prepared dataset.
   • Adjust hyperparameters such as learning rate, batch size, and number of epochs as needed.
   • Execute the notebook cells to train the model.

3. Model Evaluation:

   • Evaluate the trained model's performance on a separate test set.
   • Use metrics like accuracy, precision, recall, and F1-score to assess the model's effectiveness.

4. Model Deployment:

   • Once satisfied with your trained model, save it.
   • Update the app.py file to load and utilize your custom-trained model.

Acknowledgments:

• This project was inspired by the Galaxy Zoo challenge on Kaggle.
• We thank the creators of EfficientNet and PyTorch for their amazing work.