Results & Review

1. 1Browse and filter: Use the search box ("Search by number, team, or filename...") to find a photo, or the filter dropdown to narrow the grid to All Results, Matched Only, ⚠️ Needs Review, No Match Only, Manually Corrected, High Confidence (>90%), Low Confidence (<70%), or No Metadata Written. Clear Filters resets both.
2. 2Open a photo: by clicking its thumbnail. The counter (e.g. "101 / 149") shows your position in the batch; ←/→ (or the on-screen arrows) move to the previous/next photo without closing the view.
3. 3Zoom and pan: with the on-screen − / 100% / + / fit-to-view controls, the mouse wheel, drag-to-pan, or double-click to zoom in; the detection overlay stays glued to the image while you zoom.
4. 4Read the overlay: Detected boxes are drawn on the photo: a recognized subject shows its number/name and confidence, a detected-but-unread subject is labelled "not read", and an unmatched or too-small face is labelled accordingly. Press B (or use the gallery toggle) to show/hide the overlay; a chip reports how many results have no box to draw. The setting persists per device (Settings) and can also be turned off for the session only.
5. 5Correct a detection: Click into the Race Number or Drivers field and type; both fields save the correction and advance to the next photo on Enter. Space jumps straight into the Race Number field. Shift+Enter goes back one photo from any field. Edits made while moving between multiple detections in the same photo (↑/↓) are auto-saved as you move off each entry — no explicit Save click needed.
6. 6Add a missed detection: Every photo always shows one trailing empty entry at the end of its detection list — type a number into it to create a new detection (a fresh empty entry then appears after it). Move between entries with ↑/↓.
7. 7Delete a detection: Click its trash icon — it is removed immediately with no confirmation dialog; a toast appears with an Undo button to put it back.
8. 8Resolve an ambiguous match: A photo flagged ⚠️ Review shows a candidate panel with the tied alternatives, each with its car/vehicle model shown so you can pick the right one at a large event. With the keyboard, digits 1–9 pick the candidate at that position and 0 means "none of these" (only active on needs-review photos).
9. 9Open the shortcuts helper: Click ⌨ Want to review faster? · Discover the keyboard shortcuts, or press ?, for a cheat-sheet of every key above, grouped by Move around / Edit the number / Pick a suggestion / See what the AI saw / Done. Esc (or the backdrop/✕) closes it.
10. 10Act on "Useful Data Found.": When you click DONE! or Start New Analysis, RaceTagger may open a modal proposing preset improvements: per-participant suggestions and, at the top, a dedicated 🏁 Championship / series sponsors section (pre-checked when a sponsor covers ≥70% of the event's cars) so a series-wide sponsor like a tire or oil brand stops being matched to every car. Update Preset applies your selections; Skip (or ✕, or clicking outside) dismisses it for that analysis only — it will not auto-pop again for the same execution, but the ✨ Improve Preset button reopens it on demand.
11. 11Export or organize: 📤 Export & IPTC renames, organizes into folders and/or writes IPTC metadata (see the Folder Organization and Metadata/IPTC pages); the Deliver button sends photos to a client gallery and can upload HD originals. ✅ DONE! returns to the home screen; New Analysis starts a fresh batch.