Galaxy Classifier

This code uses PyTorch to build and implement a MLP model with PyTorch for solving a classification problem. Our goal is to classify galaxy images into 4 classes: ellipticals, lenticulars, spirals, and irregulars.

Data Documentation

We will use EFIGI dataset which contains 4458 images.

Code Explanation

• Importing libraries: The first cell imports the necessary libraries for the project, such as PyTorch, NumPy, Matplotlib, and scikit-learn. These libraries provide tools for data manipulation, visualization, and machine learning.
• Loading data: The second cell loads the galaxy images and labels from a folder using the torchvision.datasets.ImageFolder function. It also applies some transformations to the images, such as resizing, cropping, converting to grayscale, and normalizing. The data is then split into training, validation, and testing sets using the torch.utils.data.random_split function.

• Defining MLP model: The third cell defines a multilayer perceptron (MLP) model using the torch.nn.Module class. The structure of the MLP constructed in this code is as follows:

  • Convolutional and MaxPool layers: These layers extract features from the input images by applying filters and reducing the spatial dimensions. The code defines four convolutional layers and three max-pooling layers. The number of filters increases from 32 to 128 as the depth of the network increases. The kernel sizes of the filters decrease from 6 to 3 as the spatial dimensions of the images decrease. The max-pooling layers have a kernel size of 2 and a stride of 2, which means they reduce the image size by half in each dimension.
  • Dense layers: These layers perform classification based on the extracted features. The code defines three dense layers. The first dense layer flattens the output of the last convolutional layer into a vector of size 12822. The second dense layer reduces the dimensionality from 128 to 64. The third dense layer outputs a vector of size num_classes, which is 4 in this case. The dense layers use ReLU activation functions.
• Training MLP model: The fourth cell trains the MLP model using the torch.optim.Adam optimizer and the torch.nn.CrossEntropyLoss criterion. It also tracks the training and validation accuracy and loss using the sklearn.metrics.accuracy_score function. It plots the learning curves using Matplotlib.
• Evaluating MLP model: The fifth cell evaluates the MLP model on the test set using the sklearn.metrics.classification_report function. It also displays some sample predictions and their corresponding labels using Matplotlib.