Intelligent Document System

Project Overview This project comprises a production-ready, enterprise-grade Intelligent Document Processing and Classification Vault (IPCV) specifically designed for Chartered Accountant (CA) firms and financial institutions. The system automates the ingestion, classification, data extraction, validation, and segregation of complex financial documents (such as Tax Invoices, Utility Bills, GST Statements, and PAN Cards) with high-speed parallel batch processing and a modern analytical web interface. The codebase is fully functional, structured, and optimized, combining hybrid rule-based keyword matching with scikit-learn machine learning classifiers and multi-engine OCR technology (Tesseract and EasyOCR) to achieve high accuracy and eliminate data hallucinations. Technical Architecture & Core Technologies Backend Framework: Flask (Python 3.8+) Real-Time Communications: Flask-SocketIO (with Eventlet/Gevent support) Interactive Analytics Dashboard: Dash (Plotly) integrated into Flask OCR Engines: Hybrid Engine (Tesseract OCR + EasyOCR) with OpenCV image preprocessing Classification Engines: Hybrid Machine Learning (scikit-learn Random Forest/Decision Trees) + Regex & Keyword-based disambiguation Database: SQLite with SQLAlchemy ORM (for user authentication and system states) Security & Encryption: bcrypt (for password hashing), AES-256 (for output CSV/Excel reporting security) Frontend: Modern, responsive dashboard design utilizing HTML5, CSS3 (glassmorphism design aesthetic), and JavaScript (vanilla interactive components) Key Features & Capabilities 1. Robust Document Ingestion & Parallel Processing Multi-threaded and multi-process batch uploading supporting images (JPEG, PNG, BMP, TIFF), PDFs, and compressed ZIP archives. Parallel Processing Pool Executor utilizing worker initializers to eliminate startup overhead and redundant package loading, boosting throughput for large batches. 2. Hybrid OCR & Advanced Image Preprocessing Dynamic switching between Tesseract (for speed) and EasyOCR (for complex layouts and handwriting) based on confidence thresholds. Image preprocessing pipeline including auto-skew correction, adaptive thresholding, grayscale conversion, and contrast enhancement (CLAHE) using OpenCV. 3. High-Precision Hybrid Classification Dual-layer classification matching a ML classifier model (scikit-learn) with a weighted keyword registry. Smart disambiguation rules to separate invoices from utility bills (electricity, gas, water, internet, phone, credit cards, insurance). 4. Anti-Hallucination Validation Engine Strict data validation against patterns (e.g., GSTIN, PAN numbers, dates, amounts) defined by regex and business logic rules. An anomaly detection framework analyzing empty fields, repeated characters, spacing anomalies, and confidence levels to trigger human-in-the-loop review alerts. 5. Secure Output & Reporting Vault Automated document segregation into type-specific output directories. Generates comprehensive, formatted Excel reports and CSV files, automatically encrypted using AES-256 for maximum security. 6. Modern Analytics Web Interface A premium, responsive dark-themed sidebar dashboard featuring real-time log streaming, progress bars, interactive configurations, and single-click file downloads. An embedded Plotly Dash dashboard visualizing processing metrics, vendor spend, classification confidence distributions, and ROI hours saved.

Project ID: 40467784