Browse categories
Explore
Fiverr Pro
English
$
USD
I will extract data from scanned pdfs, invoices and images into excel with python ocr
Jacob Sandler
About this gig
Stop retyping. Point a script at a folder of scanned PDFs, invoices, receipts, forms or photos and get a clean spreadsheet out.
I build OCR and image-processing pipelines in Python (Tesseract, docTR, EasyOCR, OpenCV), after over a decade building imaging and inspection tools as an engineer at Johns Hopkins APL.
WHAT YOU GET
Python source, run instructions, and testing on your own samples. Low-confidence reads are flagged so you know what to check.
PACKAGES
One document type: one layout, for example one vendor's invoice, up to 5 fields, Excel or CSV output.
Batch pipeline: drop files in a folder, get one spreadsheet back. Up to 15 fields, flags for low-confidence reads, setup docs.
Vision tool: the pipeline plus a simple review window, or image tasks like detecting, measuring, counting or cropping, with an accuracy report.
Before ordering, send 10 to 20 representative files and a filled-in example of the spreadsheet you want. Accuracy depends on scan quality; I tell you what to expect after seeing your samples.
Not included: handwriting-heavy documents, custom model training and hosting.
Get to know Jacob Sandler
Jacob Sandler
Photo Editor and Python Automation Developer
- FromUnited States
- Member sinceApr 2024
Languages
English
Photographer and Python developer. I edit galleries, and I build the tools that process them.
Weddings, headshots, real estate and drone - I edit the way the photographer would have, because I am one. Skin that looks like skin, color that holds up, a gallery that looks like one shoot.
I also build software. Over a decade as an engineer at an aerospace and defense lab, writing Python and computer vision tools. Need a tool that does not exist yet? That is the other half of what I do.
Lightroom, Photoshop, Capture One. Python, OpenCV, PyQt.
Full time, fast replies, and I hit my dates.
FAQ
How accurate will it be?
It depends entirely on your images - lighting, focus, consistency and how distinct the thing is. I will not guess a number to win the order. Send samples first and I will give you a realistic range, and tell you if it is not worth doing.
Do you train custom models, or is this classical CV?
Most jobs need no training at all. OCR plus rule-based OpenCV handles printed documents and consistent images very well, and that is what every package covers. Custom model training and labeling are quoted separately after I review your samples.
My images are confidential. Is that a problem?
No. I have spent my career working under strict handling rules. I will sign an NDA, work only from what you send, delete everything on request, and I do not reuse client images in my portfolio without written permission.
