The floor report lists every page the survey read and every page OCR read (CSV), the OCR text of each file, and the full numbers with each session’s cost (JSON). Choose where it goes with Browse; without a chosen folder, your browser asks before saving it to its downloads. Your PDFs are never changed or uploaded.
Each worker opens the file in its own thread and reads every page: its text, whether that text is tagged, whether it is a scan that needs OCR or was OCR’d before, its pictures (decoded for real), and its structure tags. That is the first station of the factory, done for real here.
| Worker | Thread | Now | Pages | Pages per Minute | Time per Page |
|---|
| Session | Started | Workers | OCR Lanes | Pages | OCR’d | Time | Session Cost | Cost per Page |
|---|
Every file, page, worker and number on this floor is real: the PDFs come from your computer, each survey worker and each OCR lane is a real background thread in your browser, and the speeds are measured as the pages are read. Survey and OCR run here; OCR reads with PaddleOCR PP-OCRv6 and checks doubtful lines with Tesseract 5. Rebuilding, verifying and compiling need the TourGuid app; their counts show the work the survey found for them.