Loading...Loading...Loading...Loading...Real status of the video-to-PGN stack. Accuracy is only shown when it is measured from labelled data; untrained modules stay explicit.
not measured
not measured
Global read before opening model details.
needs more data
needs more data
Real status of the video-to-PGN stack. Accuracy is only shown when it is measured from labelled data; untrained modules stay explicit.
Baseline in use: pieces are detected with CameraChess 480M, but it is not trained on your videos yet.
Manual calibration with a8/h8/a1/h1 is the source of truth. The corner model is only a fallback.
Baseline in use: clock is read with Tesseract and corrections; there is no own 7-segment model yet.
No checklist yet.
No checklist yet.
No trained versions yet. Upload a video, calibrate board/clock, paste the correct PGN, save the dataset, then create a snapshot from admin.