Classification of Mars Lineament and Non-Lineament Structure Based on ResNet50

This article will give a brief introduction about training ResNet to classify the non-lineament and lineament structure on Mars surface, which is specific referring to crater and lineament structure which are dorsum and vallis. We build a dataset containing 300 samples of each category and splitting them into training set and test whose proportion is 4:1. After the network structure and parameters optimization, it accuracy is able to reach around 85% (for crater), 92% (for dorsum) and 80% (for vallis) respectively.