SIH OCR Studio

High-Performance Windows Desktop Application for Offline Document OCR & Text Extraction
Powered by JavaFX 21, Python 3.12, Tesseract OCR v5, PyMuPDF, OpenCV, and FastAPI.

SIH OCR Studio provides an offline desktop experience for converting scanned images and multi-page PDF documents into editable, structured text with full support for Hindi (Devanagari), English, and bilingual Hindi+English documents.

📸 Key Capabilities & UI Layout

SIH OCR Studio implements a clean and high-performance desktop architecture:

+-----------------------------------------------------------------------------------------------+
|  SIH OCR | Studio          [Engine: Tesseract OCR] [Engine Loaded ✓] [Lang: Hindi + English]  |
|  [📁 Open File]  dh2efile.pdf   [< Prev] [Page: 1/12] [Next >]  [Zoom +/-] [Fit] [Rotate]     |
+-----------------------------------------------+-----------------------------------------------+
|                                               |  Output: [Markdown] [HTML] [JSON]  [Copy]     |
|              DOCUMENT PREVIEW                 |  -------------------------------------------  |
|                                               |  OCR Result                                   |
|   • High-fidelity rendering (PyMuPDF)         |  File: dh2efile.pdf                           |
|   • Interactive zoom, pan & rotation          |  Page: 1/12 | Lines: 18 | Words: 140          |
|   • Drag-and-drop file upload                 |  Confidence: 94.8%                            |
|   • Image & PDF text-layer inspection         |  -------------------------------------------  |
|   • Multi-page navigation                     |  तमिलनाडु की राजनीति में सरकार गठन को         |
|                                               |  लेकर हलचल तेज़ हो गई है...                   |
+-----------------------------------------------+-----------------------------------------------+
|  [▶ Run OCR]  [⚡ OCR All Pages]  [⏹ Stop]     |  [Copy Text] [Export DOCX] [Export TSV]       |
|  Time: 1.4s | Progress: [==========] 100%     |  [Export TXT] [Export JSON] [Export HTML]     |
+-----------------------------------------------+-----------------------------------------------+
|  Ready | Backend: Online ✓ (127.0.0.1:8000)                                                   |
+-----------------------------------------------------------------------------------------------+

Core Features:

Offline & Private: All processing occurs locally on 127.0.0.1. No internet access required. No data leaves the machine.
Multilingual OCR Engine: Native Tesseract OCR integration with multi-script support:
- English (eng)
- Hindi (hin)
- Hindi + English (hin+eng)
- Dynamic detection of 160+ installed Tesseract language packs.
Multi-Page PDF Processing:
- Fast page navigation and progressive on-demand page rendering at 150-300 DPI.
- Hybrid text-layer detection: instantly extract embedded digital text or perform full OCR.
- Sequential batch processing with cancellation checkpoints.
Image Preprocessing Pipeline (OpenCV & Pillow):
- Automatic deskewing (rotational alignment).
- Noise reduction (bilateral filtering).
- Contrast enhancement (CLAHE).
- Adaptive thresholding (binarization).
- Upscaling for fine/small typography.
- Invert colors mode for dark background documents.
Output Views & Formats:
- Markdown Tab: Clean readable plain text with preserved paragraphs and Devanagari Unicode.
- HTML Tab: Structured, safe semantic HTML output.
- JSON Tab: Word-level positions, line indices, bounding boxes, and confidence statistics.
Export Formats:
- Microsoft Word (DOCX)
- Tab-Separated Values (TSV)
- Plain Text (TXT)
- Structured JSON
- Webpage (HTML)
- Markdown (MD)

🛠️ Technology Stack

Layer	Component	Description
Frontend UI	JavaFX 21	Native desktop UI, GPU-accelerated rendering, Segoe UI & Devanagari typography.
Build System	Maven 3.9	Dependency management, JavaFX packaging, shaded fat JAR creation.
Backend Engine	Python 3.12 (FastAPI)	High-throughput asynchronous local service bound strictly to `127.0.0.1`.
OCR Engine	Tesseract OCR v5.5	Industrial OCR engine with neural network (LSTM) recognition.
PDF Processing	PyMuPDF (fitz)	High-fidelity rendering and vector text-layer extraction.
Image Pipeline	OpenCV & Pillow	Geometric correction, filtering, and thresholding.
Export Service	python-docx & csv	Native Microsoft Word DOCX and TSV generation.

🚀 Quick Start on Windows

1. One-Click Setup

Open PowerShell in the workspace root and run:

Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
.\scripts\setup_windows.ps1

2. Launch the Desktop Application

Double-click run.bat or run in Command Prompt / PowerShell:

.\run.bat

Or via PowerShell script:

.\scripts\run_windows.ps1

3. Build & Package

To produce a standalone Windows package:

.\scripts\build_windows.ps1

🔌 Local API Endpoints (`127.0.0.1:8000`)

The application uses an internal REST API for decoupled Java-to-Python communication:

GET /api/health: Health status, Tesseract availability, version, and language pack check.
GET /api/languages: Dynamic list of available installed OCR languages.
POST /api/document/info: Document inspection (type, format, dimensions, page count, text-layer presence).
GET /api/document/preview: Render document page preview at specified DPI (PNG stream).
POST /api/ocr: Submit OCR job (single page, multi-page, or text-layer auto).
GET /api/jobs/{job_id}: Track progress (0.0 to 1.0, current page, elapsed time).
GET /api/jobs/{job_id}/results: Retrieve completed OCR text, stats, and words metadata.
POST /api/jobs/{job_id}/cancel: Request safe cancellation between pages.
POST /api/export: Export results to DOCX, TSV, TXT, JSON, HTML, or MD.
POST /api/settings/update: Dynamically update and re-test Tesseract executable path.

🧪 Automated Test Suite

Both backend and frontend components have comprehensive automated test suites:

1. Backend Pytest Suite

Run with:

.venv\Scripts\python.exe -m pytest backend/tests

Results: 8 passed in 4.71s

test_health_endpoint
test_languages_endpoint
test_document_processor_inspect
test_pdf_rendering
test_image_preprocessing
test_ocr_engine_execution
test_api_ocr_flow
test_exports (TXT, DOCX, TSV, JSON, MD, HTML)

2. Java Frontend Unit Tests