This project focuses on developing an Optical Character Recognition (OCR) system for recognizing both handwritten and printed Malayalam text from scanned PDF documents. The work is split into two parts:
- OCR for Handwritten Malayalam Characters
- OCR for Printed Malayalam and English Text
This module focuses on recognizing individual Malayalam characters from handwritten inputs. Key stages in this pipeline include:
- Document Preprocessing
- Text Segmentation
- Character Recognition using Deep Learning
- Achieved 91% accuracy at the character level.
- Word/sentence recognition is feasible only with consistent spacing.
- Built a minimalistic and scalable OCR system for handwritten character-level recognition.
- Paves the way for digitization and automated data entry from handwritten documents.
This module deals with the recognition of printed Malayalam (and English) text from scanned PDFs using a hybrid approach combining:
- Tesseract OCR
- Deep Learning-based Preprocessing Techniques
- PDF to Image conversion
- Grayscale and binary image preprocessing
- Multi-threaded and language-optimized OCR execution
- Parallelized OCR pipeline to enhance performance
- Improved recognition accuracy over traditional OCR tools
- Effective for multilingual documents
- Python
- OpenCV
- Tesseract OCR
- TensorFlow / Keras
- Pillow
- NumPy
- PDF2Image
- Matplotlib (for visualization)
- Robust recognition of handwritten Malayalam characters
- Dual-language OCR system for printed Malayalam and English
- Optimized for document digitization workflows
- Scalable and minimalistic system architecture
📦 CDIT_Team2_Internship/
├── 🌿 handwritten_ocr/ (branch: `main`)
│ ├── 📁 Code/
│ ├── 📁 Dataset/
│ ├── 📁 model/
│ ├── 📁 Testing the Model/
│ └── 📄 README.md
└── 🔀 printed_ocr/ (branch: `Printed-Documents`)
├── 📄 main.py
└── 📄 README.md