Mars Landmark Classifier
Classifying Mars landmarks from compressed encodings
- 84.26%test accuracy classifying from the encodings
- 98.4%dimensionality reduction, 224×224 images to 500-length vectors
I led a team building a Mars landmark classifier around a custom CNN autoencoder. The encoder compresses 224×224 images into compact 500-length latent vectors, a 98.4% reduction in dimensionality, and classification runs on those encodings rather than on raw pixels.
To balance the dataset, I generated 2,000 synthetic images for underrepresented classes using augmentation.
The model reached 98.98% training accuracy and 84.26% test accuracy. The gap between the two points to overfitting, which is the clearest place to improve next.