The Mars Landmark Classifier is a system designed to enable efficient data transmission between space rovers and mission control. By incorporating an image compression pipeline and a classification model, this project addresses the unique challenges of space communication while preserving critical image data for scientific analysis.
- Image Compression System:
- Utilized a CNN AutoEncoder to compress 224x224 resolution images by 99%, reducing them to 500-length latent vectors while maintaining image quality.
- Mars Landmark Classification:
- Achieved 84.26% accuracy on decoded feature vectors using a CNN-based classifier.
- Dataset Expansion:
- Augmented the dataset with 2000 synthetic images to address class imbalance, improving model robustness and performance.
- Validated the effectiveness of the compression system by achieving high classification accuracy on decoded vectors.
- Significantly optimized data transmission for Mars missions, reducing communication overhead while preserving actionable data.
- TensorFlow for deep learning model development.
- Keras for implementing the CNN AutoEncoder and classifier.
- Image Augmentation techniques to expand the dataset and balance classes.
- Clone the repository:
git clone https://github.com/yourusername/mars-landmark-classifier.git cd mars-landmark-classifier