Skip to content

Latest commit

 

History

History
64 lines (49 loc) · 2.74 KB

README.md

File metadata and controls

64 lines (49 loc) · 2.74 KB

Galaxy-ZOO-Image-Classifier-with-Deep-Learning

Galaxy Image Classifier

Overview

This project uses a Convolutional Neural Network (CNN) to classify galaxy images into five distinct categories based on the Sloan Digital Sky Survey (SDSS) data. The images are preprocessed and trained using a CNN model to achieve a high classification accuracy of 89.4%.

Data Source

The galaxy classification dataset was obtained from the following sources:

Galaxy Classes

The dataset consists of five galaxy types:

  • Cigar-shaped smooth: Class 0
  • In between smooth: Class 1
  • Completely round smooth: Class 2
  • Edge-on: Class 3
  • Spiral: Class 4

Methodology

Techniques Used

  1. Data Preprocessing:

    • Loaded image paths and labels from a CSV file.
    • Verified image paths and filtered invalid entries.
    • Normalized image pixel values to the range [0, 1] using ImageDataGenerator.
    • Resized images to 128x128 pixels.
    • Split the dataset into training (80%), validation (5%), and testing (15%) subsets.

  2. Model Architecture:

    • Built a sequential CNN with three convolutional layers, each followed by max-pooling.
    • Flattened the output of the convolutional layers and added fully connected dense layers.
    • Included a dropout layer (rate = 0.5) to prevent overfitting.
    • Used a softmax activation function in the output layer for multi-class classification.

  3. Training and Evaluation:

    • Used adam optimizer with categorical cross-entropy as the loss function.
    • Trained the model for 10 epochs with batch size 32.
    • Visualized accuracy and loss trends for training and validation datasets.

  4. Performance Metrics:

    • Calculated a test accuracy of 89.4%.
    • Analyzed the model's performance using a confusion matrix and classification report.

Model Summary

  • Input Shape: (128, 128, 3)
  • Convolution Layers:
    • Conv2D with 32, 64, and 128 filters, each with kernel size (3x3) and ReLU activation.
    • MaxPooling2D after each convolutional layer.
  • Fully Connected Layers:
    • Dense layer with 128 units and ReLU activation.
    • Dropout layer with a rate of 0.5.
    • Output layer with 5 units (softmax activation).

Visualization

  • Confusion matrix and classification report highlight model performance.
  • Accuracy and loss trends show consistent improvements over training epochs.

Results

The CNN achieved an impressive accuracy of 89.4%, demonstrating its effectiveness in classifying galaxy images into their respective categories.