2025 workshop program
Exospore: AI-Driven Detection of Microbial Life for Planetary Protection
The work
Abstract
Preventing forward contamination during interplanetary missions is a critical challenge for planetary protection. Exospore is a collaborative research effort between the SETI Institute, NASA Ames, and the University of Florida, employing deep learning to detect bacterial spores in scanning electron microscope (SEM) images of spacecraft-relevant materials. By leveraging Faster R-CNN architectures optimized with large-scale SEM data and expert-labeled annotations, Exospore enhances sensitivity to rare and morphologically ambiguous spores—potential threats for cross-planetary microbial transfer.
This talk presents our data pipeline, from high-fidelity sample preparation and SEM imaging to multi-institutional annotation and model training using Google Cloud infrastructure. Results demonstrate robust detection across varying contamination contexts, with performance comparisons between backbone architectures (e.g., X101-FPN vs R101-C4). Our work advances AI applications in astrobiology by scaling the detection of trace biosignatures in extreme datasets, directly contributing to NASA’s planetary protection protocols.