What can be observed
Numbered open circles span the printed Tartu street map. Labels 1–81 are visible after rectification and close inspection.
Interactive evidence explorer
Follow the whole chain—from a photographed map to route scenarios.
Inspect the original sheet, verify all 81 detected circle centres, review the disputed geographic transform, then compare best-found walking routes under different pace and time limits.
81source detections
—frontier routes
4–42 kmsearched budgets
01 · Original evidence
The unmodified photographed map
This is the exact image supplied in Discord. The sheet is photographed vertically with perspective, curved paper, fingers at two corners, and a complete printed control map visible across the frame.
1552379705661853728Why rectification is needed
The paper is rotated, curved and viewed obliquely. Detection coordinates are therefore measured on a perspective-rectified image, not directly on the camera pixels.
What the app preserves
The exact source image, asset hashes, all 81 rectified circle centres, and the later retraction are frozen together so the evidence chain remains auditable.
02 · Supported layer
All 81 detected circle centres
These source-image centres and visible labels remain supported. Select any control to inspect its exact rectified pixel location and cross-highlight it on the provisional map.
Supportedimage-space centres + labels
03 · Disputed layer
The provisional geographic mapping
The marker map makes the disputed transformation inspectable rather than hiding it. Select markers and compare them with the source detections above.
Retractedtransform + WGS84 coordinates
Unverified · not for navigation
04 · Experimental route lab
Compare best-found routes by pace and allowed time
Change either constraint. The app converts them into a maximum walking distance and selects the lexicographically best precomputed candidate: most controls first, then shortest distance.
Experimentalroutes inherit disputed coordinates
Unverified · not for navigation
Best found · not proven optimal
Selected candidate—
Search source—
Route frontier
Controls gained as distance increases
Visit order
Selected route
Preset comparison
Several constraints, one route frontier
Method & provenance
What each layer means
Source observations
Original photo, rectified image, 81 visible labels and their source-image circle centres.
Geographic interpretation
Image-to-reference transformation, WGS84 coordinates and claims about which real-world road, bank or junction contains each control.
Route scenarios
Pedestrian shortest paths and prize-collecting routes derived from the retracted coordinates. Useful for exploring the optimization model only.
Technical route model
Fixed start: University of Tartu Old Observatory. Fixed finish: Püssirohukelder. For each distance budget, optional controls are optimized over an OpenStreetMap pedestrian shortest-path matrix. The objective is lexicographic: maximize visited controls, then minimize route distance. The static frontier combines three RoutingModel strategies at 4–42 km plus the extended 49-control and exact-50 searches. Results are “best found”, not globally proven.
Evidence provenance
Source photo message: 1552379705661853728. Refreshed 81-control extraction message: 1552404436729471037. Retraction message: 1552459644243550350. The app ships exact frozen source assets and hashes in its data bundle.