ExoMinerNEW

ExoMiner is a deep neural network classifier developed by NASA specifically for the automated validation and classification of exoplanet candidates. The software analyzes photometric light curves and diagnostic data from space telescope missions such as Kepler, K2, and TESS. Using machine learning algorithms, the system accurately distinguishes genuine extrasolar planets from false-positive signals, such as eclipsing binary stars.

The system processes complex time-series and spatial diagnostic data, including centroid shifts and pixel-level metrics, to verify transit signals. ExoMiner combines domain knowledge from exoplanetary science with modern deep learning architectures, achieving high classification accuracy. In scientific studies, the model successfully validated hundreds of new exoplanets in Kepler's archival data that had previously gone undetected.

As an open-source project hosted on GitHub, ExoMiner is freely available to scientists, astrophysicists, and data researchers worldwide. Users can run the Python code locally, train it on custom datasets, or adapt the neural network architecture to new observational surveys. ExoMiner provides a vital tool for automating the analysis of massive astronomical datasets.

• Deep Learning Classification — Utilizes deep neural networks to accurately distinguish planetary transits from false-positive signals. Processes photometric time series alongside diagnostic parameters from Kepler and TESS missions.
• Multi-Feature Diagnostic Pipeline — Evaluates light curve dips together with centroid motion and pixel-level information. Significantly reduces the rate of false positives compared to traditional heuristic filters.
• Open-Source & Scalable — The source code is fully transparent and structured for execution on GPU-enabled hardware and HPC clusters. Researchers can customize and extend the model for dedicated scientific surveys.

ExoMiner is designed primarily for astrophysicists, astronomical researchers, and data scientists working in planetary science. It is also used by academic institutions and space science research labs processing large space telescope datasets.

An astrophysics research group is analyzing terabytes of light curve data from the NASA TESS mission to discover potential exoplanet candidates. Instead of manually inspecting thousands of signal dips, the team deploys ExoMiner on a local GPU cluster. The model rapidly filters out background eclipsing binaries and instrument noise, generating a highly reliable list of validated exoplanet candidates for telescope follow-up observations.

• ExoMiner is designed primarily for astrophysicists, astronomical researchers, and data scientists working in planetary science. It is also used by academic institutions and space science research labs processing large space telescope datasets.
• An astrophysics research group is analyzing terabytes of light curve data from the NASA TESS mission to discover potential exoplanet candidates.
• Deep Learning Classification: Utilizes deep neural networks to accurately distinguish planetary transits from false-positive signals. Processes photometric time series alongside diagnostic parameters from Kepler and TESS missions.
• Multi-Feature Diagnostic Pipeline: Evaluates light curve dips together with centroid motion and pixel-level information. Significantly reduces the rate of false positives compared to traditional heuristic filters.
• Open-Source & Scalable: The source code is fully transparent and structured for execution on GPU-enabled hardware and HPC clusters. Researchers can customize and extend the model for dedicated scientific surveys.

ExoMiner should not be used for general-purpose time-series forecasting outside astronomy or standard business data analytics. It is also unsuitable for non-technical users lacking Python programming experience and background in transit photometry.

• ALeRCE — Free · HQ: Santiago, Chile · Unclear
• AstroLens — Free · HQ: United States · GDPR / EU
• Astrometry.net (Nova) — Free · HQ: New York, United States · Unclear
• Galaxy Zoo — Free · HQ: Oxford, United Kingdom · Unclear
• GraXpert — Free · HQ: Germany · GDPR / EU