Windows App: Scan Tables to Excel
Budget: €30 – €250 EUR
I have about one hundred A4 sheets, each printed with a seven-column, roughly thirty-row table that has been filled in by hand (mostly names and numbers). I need a small Windows desktop application that will let me run those pages through a scanner and end up with a clean, validated Excel workbook.
Core workflow
1. From the app’s interface I choose a batch of PDFs or images coming straight from the scanner.
2. The app sends each page to an online handwriting-recognition service—whichever provider gives reliable results at the lowest cost is fine (Google Cloud Vision, Azure Cognitive Services, Amazon Textract or another option you can recommend).
3. The returned text is mapped back to the seven fixed columns and written to an .xlsx file.
4. Before the rows are finally saved, basic data validation is applied: empty required cells flagged, obvious type mismatches corrected or surfaced, and duplicate names highlighted.
5. I can review a simple on-screen report of any rows that need manual attention, then export the clean sheet.
Deliverables
• Windows-only desktop executable (no command-line usage required).
• Source code plus brief setup instructions.
• An .xlsx file produced from a small sample stack I will supply to prove accuracy and validation logic.
Acceptance criteria
– At least 95 % character recognition accuracy on my sample pages.
– Seven columns appear in the exact order of the original form.
– Validation catches blanks, non-numeric entries in numeric columns, and duplicate names.
– App runs on any current Windows 10/11 machine without additional paid licences beyond the chosen AI API.
I value a straightforward, lightweight solution—no unnecessary bells and whistles—just a dependable way to turn these handwritten tables into clean Excel data at minimal API cost.
Core workflow
1. From the app’s interface I choose a batch of PDFs or images coming straight from the scanner.
2. The app sends each page to an online handwriting-recognition service—whichever provider gives reliable results at the lowest cost is fine (Google Cloud Vision, Azure Cognitive Services, Amazon Textract or another option you can recommend).
3. The returned text is mapped back to the seven fixed columns and written to an .xlsx file.
4. Before the rows are finally saved, basic data validation is applied: empty required cells flagged, obvious type mismatches corrected or surfaced, and duplicate names highlighted.
5. I can review a simple on-screen report of any rows that need manual attention, then export the clean sheet.
Deliverables
• Windows-only desktop executable (no command-line usage required).
• Source code plus brief setup instructions.
• An .xlsx file produced from a small sample stack I will supply to prove accuracy and validation logic.
Acceptance criteria
– At least 95 % character recognition accuracy on my sample pages.
– Seven columns appear in the exact order of the original form.
– Validation catches blanks, non-numeric entries in numeric columns, and duplicate names.
– App runs on any current Windows 10/11 machine without additional paid licences beyond the chosen AI API.
I value a straightforward, lightweight solution—no unnecessary bells and whistles—just a dependable way to turn these handwritten tables into clean Excel data at minimal API cost.