Back to Projects Library
Computer Vision & OCRCompleted2026
ShreeAkshara: Computer Vision for Indic Manuscripts
Role: Research Intern (IIT Tirupati)
Computer vision research project on Indic script style adaptation and augmentation for low-resource Devanagari text recognition.
Verified Project Links & Repositories
1 Active ResourceTechnology Stack & Libraries
PythonOpenCVPyTorchCLAHEHough TransformTesseractEasyOCR
The Challenge / Problem
Context & Objectives
Historical Indic manuscripts suffer from ink bleed, degradation, and non-linear baselines that defeat standard OCR tools.
Architectural Solution
Engineered Approach
Developed CLAHE preprocessing, Hough line extraction, and synthetic degradation datasets to improve transcription robustness.
System Architecture & Pipeline
Degraded Image -> CLAHE Preprocessing -> Hough Line Segmentation -> OCR Transcription -> IEEE Report.
Contributed to ShreeAkshara, building a multi-module computer vision pipeline for palm-leaf manuscript analysis with Hough line segmentation, CLAHE enhancement, and synthetic style augmentation.
Engineered Capabilities & Innovations
Hough-transform line segmentation for curved manuscript text lines
CLAHE contrast enhancement tuned for fragile palm-leaf textures
Evaluation of OCR models (Tesseract, EasyOCR) against degraded specimens
Synthetic style augmentation pipeline for historical Indic scripts
Empirical Results & Benchmarks
Authored full IEEE-format research report and documentation for Vision Technology Lab, IIT Tirupati.
Have Questions About This System?
I am always glad to discuss technical architecture, benchmarks, or potential collaboration.