OCR Using Vision Transformer (ViT)

The coode is working on the ocr of indic languages which is done by usingf thge technology nad concepts of transformers the deep learning ways natural language pros\cessing made it a sucess.This repository demonstrates an end-to-end Optical Character Recognition (OCR) system using the Vision Transformer (ViT) architecture. It focuses on recognizing characters from images using a deep learning model implemented in PyTorch.

---

🚀 Overview

Traditional OCR systems often use CNN-based architectures. This project leverages the transformer-based ViT model, which segments the image into patches and processes them similarly to tokens in NLP transformers. This allows the model to capture global context more effectively.

The model is trained to classify individual characters based on their image representation.

---

🧠 Model Architecture

• Backbone: google/vit-base-patch16-224 (Hugging Face Transformers)
• Classification Head: A linear layer to map encoded features to class logits.
• Loss Function: CrossEntropyLoss
• Optimizer: Adam
• Scheduler: StepLR

---

📂 Defining Working of ViT