Move to DiGi-Organization: update owner links and workspace paths The DiGi repositories were transferred from ZiolkowskiJakub to DiGi-Organization (client projects to DigiProject-Organization) and the local workspace folder was renamed from DigiProject to DiGi. - Owner replaced in links, issue references and sync-wiki.yml (old URLs keep redirecting). - Workspace paths DigiProject/... -> DiGi/... in docs and comments. - Regenerated .agents and the README guideline block, which also carries the pending Coding - YOLO.md guideline updates. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Year Built Prediction: dating is detector-only, regressor parked as baseline The first-detection rule (first-confident-detection@0.5) is the production predictor; the shipped .mlnet is the evaluation baseline it beats on unseen counties (DiGi.GIS.ML#15). Never-detected buildings are cleared, not left stale (DiGi.GIS.YOLO.UI#28), and the run recipe now documents the tray task's forced flags. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
train9_fresh and the 2026-10-08 regressor are the shipped pair (DiGi.GIS.YOLO.UI#12, DiGi.GIS.ML#13) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Year Built: stored User year built label source, five labelled counties, the #4 bar and the held 2026-09-29 retrain (ZiolkowskiJakub/DiGi.GIS.ML#11)
Year Built Prediction: state how the stored year is chosen — most frequent within a source, and the calculated user→predicted fallback (DiGi.GIS.PostgreSQL#94)
Fix #9: the committed options defaults now carry their clean names
Add Year Built prediction guides Two hand-authored pages, linked from Home: Year-Built-Prediction covers what the model predicts, what it actually learns, the runtime and server-side requirements, and how to drive DiGi.GIS.YOLO.UI.ConsoleApp.exe - options, exit codes, resuming, and the behaviour worth expecting. Training-the-Year-Built-Model is the retraining procedure end to end, including that Model Builder cannot train this on Visual Studio 2026 and why, and that the acceptance bar is the measured baselines rather than a headline R squared. Both record the traps that cost time: the two indistinguishable 401s on the write endpoint, the label semantics, and county-specific orthophoto coverage. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>