This project compares classical machine learning and deep learning approaches for classifying Martian terrain using labeled satellite imagery, including SVM and convolutional neural network models.
Repository Structure: LICENSE - MIT open-source license README.md - project overview and usage description sample.images.mars.pdf - Sample subset of 4 images per class (8) Dataset: Full Mars terrain dataset (6,000+ images, 8 classes) available at (https://www.kaggle.com/datasets/aumthaker/mars-terrain-images?resource=download) mars.svm.py - Python script containing preprocessing, SVM model training, and evaluation mars.cnn.py - Python script containing preprocessing, CNN model training, and evaluation mars.accuracy - python script for the comparison graph mars.plots.pdf - output visualizations including confusion matrices and model comparison figures requirements.txt - Python dependencies for reproducibility