Introduction

INEC-OCR is a system that leverages Optical Character Recognition (OCR) technology for accurately extracting and verifying results from INEC statement of result forms to detect and prevent electoral fraud.

Key Features

• Automated Data Extraction: Designed and implemented an OCR pipeline to extract structured data (e.g., candidate names, vote counts, and polling unit details) from scanned images of INEC result forms.
• Data Validation: Integrated cross-verification mechanisms to detect discrepancies by comparing extracted data with official results from INEC.
• Fraud Detection: I also tried to incorporate algorithms to flag tampered forms by analyzing inconsistencies in text and handwriting patterns. This is still a work in progress.
• User-Friendly Interface: Developed an intuitive web page for extracting the results, visualizing the extracted results, and generating downloadable reports.

Technologies Used

• Python
• FastAPI
• Machine Learning: for implementing the agglomerative clustering algorithm
• OCR Engine: PaddleOCR
• Docker
• k8s

Potential Use-Cases

• Enhancing transparency and accountability in the electoral processes
• Minimizing the risk of manual errors and manipulation in INEC statement of result forms