TableOCR

Invoice & Gas Station Bill Table Extraction Pipeline

Key Result & Impact Benchmark

Automated daily financial ledger ingestion for 50+ US commercial clients with near-zero error rates.

Technical Overview

TableOCR is an automated document processing pipeline designed to parse unstructured invoices, supermarket receipts, and gas station billing sheets into structured Excel and CSV accounting formats. Combining Google Cloud Vision API, Tesseract OCR, and a custom Faster-RCNN region proposal network, the pipeline detects line items, itemized prices, tax headers, and table contours using geometric bounding box heuristics even on distorted or low-resolution scans.

Key Technical Highlights

Processed high-volume paper invoices across 50+ gas station commercial clients in the US.
Built geometric bounding-box clustering heuristics to group disconnected cells into contiguous data tables.
Reduced manual data entry overhead by over 80% with high table parsing accuracy.

Technologies & Frameworks

GCP Vision API

Tesseract OCR

Faster-RCNN

OpenCV

Python