dipanjannc/getting-started-with-ocr

Getting-Started-with-OCR

Description

Identify images based on the annotated labels provided in the samples.

Installation

pip install -r requirements.txt

Architechture

1. Image Processing

   • Orientation Correction
   • Image Binarization
   • Denoising
   • Deskewing
   • Normalization
   • Image Scaling
   • Removing Horizontal and Vertical Lines
   • Enchaining Contrast
   • Image Segmentation through boxes

2. Text Processing

   • Basic Text Cleaning (like removing repetitive special characters)
   • Spelling Correction through SYMSPELL
   • NER - For Names and Locations
   • Keyword Extraction
   • Profiling Text Data

3. Text Embeddings

   • Transformers based Embeddings
   • Sentence Transformers based Embeddings
   • FastText Embeddings

4. Model Selection & Training

5. Model Evaluation

Future Work

Some of the research findings are shared in the Document shared.