How it works?
1) Before you start
- Browser: Use a modern browser (Chrome, Edge, Firefox, Safari). Keep a single tab active when running detections.
- Data access: Orama X fetches light curves and metadata from MAST/SPOC and can query TESSCut and Gaia DR3.
- Terminology: Light curve = Flux vs time; Phase folded = folded on candidate period; BLS = Box Least Squares; Vetting = checks before confirming a planet; P(planet) = AI classifier probability.
2) Detection panel overview
The main panel lets you choose a source, configure preprocessing, run a period search, and inspect candidates. Use Fetch & Detect to download data, apply filters, and search for transit signals.
3) Single-target workflow
- Enter target ID (TIC / EPIC / Kepler) and keep Mission = auto.
- Choose preprocessing: quality mask ON, outlier σ = 5, detrend = flatten.
- Enable centroid vetting and Gaia neighbors for contamination checks.
- Set thresholds (p = 0.5–0.8, centroid σ thr = 3.0, Gaia radius = 60″).
- Click Fetch & Detect to run the search, then inspect candidates visually.
- Check metrics like SNR, ΔBIC, Odd/Even Δ, Secondary?, and Centroid.
- Fit the transit with batman, review parameters, and export vetted results.
4) Batch workflow
Analyze multiple targets using the same settings. Paste TICs/EPICs inBulk mode, configure global parameters, run sequentially, and export results (all or vetted only).
5) Custom light curves
Upload TXT/CSV files with at least time and fluxcolumns. Configure detrending and vetting as usual, then run detection and export.
6) AI / ML tools
- Classification: Each candidate receives a P(planet) score.
- Explainability: “Explain prediction” highlights top contributing features (depth, duration, SNR, etc.).
- Retraining: Upload labeled CSVs and click “Train new model” to fit a custom classifier and view metrics (F1, PR AUC, confusion matrix).
7) Gaia DR3 Neighbors panel
Shows nearby Gaia sources (sep, dx, dy, Gmag, BP−RP, RUWE). Use this to assess contamination and confirm centroid results.
8) Candidates table
Review each candidate’s period, duration, depth, power, P(planet), SNR, ΔBIC, odd/even Δ, and centroid status before confirming as a planet.
9) Export & reproducibility
Use Export CSV, Export Vetted CSV, andDownload PDF report to save your analyses. Always record mission, detrending, thresholds, and model version for reproducibility.
10) Troubleshooting & best practices
- No candidates? Increase k peaks or relax σ clip.
- Spurious periods? Tighten quality mask and verify by eye.
- Centroid fails? Reduce Gaia radius and recheck neighbors.
- AI over-confident? Raise planet threshold or review explanations.
11) Quick recipes
- Single bright TESS target: Fetch → Detect → Fit → Verify → Export.
- Batch analysis: Bulk mode → run → Export Vetted CSV.
- Custom data: Upload → Detect → Fit → Explain → Export.
12) Sensible thresholds
p = 0.5 (exploratory) / 0.8 (high purity), Centroid σ thr = 3.0, Gaia radius = 60″, k peaks = 3 (quick) / 5–10 (deep search).
Final note
Orama X enables human-AI collaboration for exoplanet discovery. Always combine AI probabilities with physical vetting (centroid, Gaia neighbors, odd/even tests, ΔBIC) and visual inspection before confirming a planet.
ORAMA X