Applications like disaster management, urban planning, and environmental monitoring rely on satellite image categorization. This project develops a machine learning pipeline using MobileNetV2, a CNN architecture, to classify high-resolution satellite images. It employs two convolutional layers (3×3 kernels) with ReLU activation, 2×2 max-pooling, a fully connected layer, and a SoftMax output for multi-class classification. Images are resized to 200×200 pixels (RGB) to balance detail and efficiency. MobileNetV2 was chosen for its low latency and high performance, using depth-wise separable convolutions and inverted residuals. The model, optimized with Adam and categorical crossentropy, achieved 98% validation accuracy and F1-scores above 0.96 across all classes, converging in 8 epochs. The architecture balances simplicity and performance for robust feature learning and generalization. This approach highlights CNNs’ ability to classify satellite images effectively. Future work could explore transformer-based models or integrate temporal satellite data to enhance analysis. This work offers a scalable, automated solution for satellite image classification.