DocRebuild AI

DocRebuild AI is a production-grade AI-powered document reconstruction platform designed to transform scanned PDFs, image-based textbooks (including NCTB books), and worksheets into highly accurate, editable, and searchable Microsoft Word (.docx) documents. Unlike traditional OCR tools that simply extract text, DocRebuild AI combines document layout analysis, multi-engine OCR, computer vision, document understanding, mathematical formula recognition, and automated quality assurance to reconstruct complex documents while preserving their original visual structure. ## Key Features • Converts scanned PDFs and image-based documents into editable DOCX files. • Preserves page layout, formatting, images, tables, captions, headers, footers, and semantic reading order. • Supports Bangla, English, multilingual documents, and mathematical equations. • Uses a multi-engine OCR ensemble with confidence-based validation and correction. • Performs AI-powered layout detection, document understanding, and visual verification. • Includes automated quality assessment, self-correction, and reconstruction validation. • Fully offline, privacy-focused architecture with optional GPU acceleration. ## Processing Pipeline PDF Ingestion → Layout Analysis → OCR Ensemble → Document Understanding → Vision Validation → Table Extraction → Math Recognition → Bangla Validation → DOCX Reconstruction → Quality Assurance → Self-Correction → Visual Verification ## Technologies FastAPI • React • TypeScript • Python • Docker • SQLite • PyMuPDF • EasyOCR • Tesseract OCR • Surya OCR • Florence-2 • Microsoft Table Transformer • Pix2Tex • OpenXML • Computer Vision • Document AI