search

Abhishek-A2077 / Classification-of-Barred-and-Unbarred-Galaxies-

This project classifies galaxies into barred and unbarred types using a VGG16-based neural network, achieving 87% accuracy. Trained on the Galaxy10 SDSS dataset, it demonstrates effective transfer learning for astronomical image classification. Scripts for training, evaluation, and testing with example predictions are included.
0 stars 1 forks source link
computer-vision hacktoberfest hacktoberfest-accepted hacktoberfest2024 space vgg16

readme

Galaxy Classification: Barred and Unbarred Galaxies

This project classifies barred and unbarred galaxies using deep learning models. Leveraging the Galaxy10 SDSS dataset, we categorize galaxies into three specific classes: Barred Spiral, Unbarred Tight Spiral, and Unbarred Loose Spiral. The models implemented include VGG16 and ResNet50, achieving notable accuracy results.

Overview

The project uses the Galaxy10 SDSS dataset to classify galaxies into three types:

These classes help distinguish between barred and unbarred galaxies for our classification tasks.

Dataset

The Galaxy10 SDSS dataset comprises images of ten galaxy classes, but only three are used for our classification task. The dataset's galaxy labels are:

  1. Disturbed Galaxies
  2. Merging Galaxies
  3. Round Smooth Galaxies
  4. In-between Round Smooth Galaxies
  5. Cigar Shaped Smooth Galaxies
  6. Barred Spiral Galaxies
  7. Unbarred Tight Spiral Galaxies
  8. Unbarred Loose Spiral Galaxies
  9. Edge-on Galaxies without Bulge
  10. Edge-on Galaxies with Bulge

Model Architecture

Two CNN models were implemented for galaxy classification:

VGG16

ResNet50

Results

Model Accuracy
VGG16 87%
ResNet50 92.8%

Setup

To set up and run the project, follow these steps:

  1. Clone the repository: 
    
git clone https://github.com/yourusername/galaxy-classification.git
cd galaxy-classification