The night sky generates terabytes of new data every night, but buried within those countless pixels are the universe’s rarest phenomena—supernovae explosions, galaxy collisions, and gravitational lenses that could reshape our understanding of cosmology. The challenge? Even dedicated survey teams can’t manually review millions of astronomical images to find these cosmic needles in the haystack.
AstroLens transforms this astronomical data deluge into discovery gold using autonomous AI that never sleeps. Built with PyTorch and Vision Transformers, it continuously ingests real sky survey data, applies YOLO object detection and out-of-distribution scoring to flag anomalies, then cross-references findings against SIMBAD and NED catalogs. During its autonomous streaming discovery run, it analyzed over 22,000 images across 5,471 sky regions and independently flagged notable objects including Supernova SN 2014J and the violent galaxy merger NGC 3690—all without being told what to look for.
This isn’t just another machine learning project—it’s a tireless research assistant that’s already proving its worth in real astronomical discovery. With FastAPI integration for real-time streaming, citizen science applications, and self-correcting algorithms that achieved 99.5% accuracy, AstroLens represents the future where AI and human astronomers work together to unlock the universe’s secrets. Whether you’re running your own sky survey or contributing to distributed astronomical research, this tool brings professional-grade anomaly detection to any Python environment.
⭐ Stars: 4
💻 Language: Python
🔗 Repository: deepfieldlabs/astroLens