Celestial Classifier: Distinguishing Earth and Mars with Neural Networks

Hi this is Immanuel and In this project, I developed a neural network-based classifier to distinguish between images of Earth and Mars. The model was trained on image datasets to learn unique visual patterns of each planet. This approach demonstrated the effectiveness of deep learning in image-based planetary classification.

Supplies

Google Colab: For building and training the neural network model.
Google Images: For sourcing training datasets of Earth and Mars images.

To train the classifier, I gathered image datasets of Earth and Mars from Google Images. The images were manually curated to ensure diversity in surface features. I then preprocessed the data by resizing all images to a uniform dimension and normalizing pixel values for consistency, making them suitable for input into the neural network.
below are the code I used.
and also I checked the images

Here I created the model, compiled, and trained using the below set of codes. I’ve added 4 layers to make the prediction better.

Since I used Mac for this project, comprising the files to upload them in Google Colab, it also adds some extra files to the model, which causes problems while predicting and compiling as the model struggled to differentiate those files(DS.store). So, I used these sets of codes to remove those files.

Finally, the model classified both Mars and Earth with an accuracy of around 98 to 99.

Since this is a simple project, I really learned the basics and the possible problems that might occur while dealing with these kinds of programs. With this, I'll continue to improve by taking more complex problems to find the optimum solutions.