Extraction MTO from scanned or image P&ID PDF

I have several P&ID drawings available only as flattened PDF images and I need an automated way to turn them into a structured Material Take Off. The workflow I have in mind relies on computer-vision / AI so that every relevant symbol is spotted, classified, and exported straight into a single Excel workbook.

Scope of recognition
• Symbols: valves, pumps, pipes, all instruments, line numbers, and any line breaks.
• Details captured per symbol: label or tag/name, quantity, and any specification noted on the drawing (size, rating, material, etc.).

Expected deliverable
An Excel file with separate, clearly labelled sheets—or a single sheet with tidy columns—containing one row per detected item. Columns must include at least: Drawing ID, Symbol Type, Tag/Name, Quantity, Specification, and the coordinates or reference for quick cross-checking on the original PDF.
Extract all VALVE ,tee ,of each line no . identify specification break.

Acceptance
1. Your extraction script or model must handle multiple P&IDs with different drafting styles without manual re-training.
2. A spot-check on three random areas of the drawing should match the Excel output 100 %.
3. All line breaks must be flagged so that downstream pipe counts reconcile.

Feel free to leverage OpenCV, Tesseract, TensorFlow, YOLO, or another approach—just keep the setup reproducible so I can run it again on future packages. A short read-me and sample code are welcome alongside the Excel file.