Aymen016 / Optical-Character-Recognition

This repository hosts the Optical Character Recognition (OCR) system, a machine learning-based project designed to recognize handwritten digits (2 and 4) using the Naive Bayes algorithm. The system processes image data, extracts features, and uses statistical models to predict the digits.

🖼️ Optical Character Recognition (OCR) System

🌟 About

This project implements a Naive Bayes-based OCR System that recognizes handwritten digits (2 and 4). The goal is to process images, extract features, and predict digits based on pixel values. It’s a simple yet effective application of machine learning techniques to solve Optical Character Recognition tasks.

🚀 How It Works

The OCR system processes images of handwritten digits, extracts pixel data, and applies the Naive Bayes classifier to predict the digit in the image. By analyzing the probability of pixels being part of specific digits, the model classifies the image based on the highest likelihood.

Key Features:

• Image Preprocessing: Converts raw pixel data into feature vectors suitable for classification.
• Naive Bayes Classifier: A powerful algorithm that leverages Bayes' Theorem to classify images based on pixel likelihoods.
• Data Handling: Handles both training and test data by splitting based on classes (2 and 4), and calculates class and feature probabilities.

📊 Model Evaluation

After training, the model's performance is evaluated based on:

• Confusion Matrix: Shows true positive (TP), false positive (FP), true negative (TN), and false negative (FN) values.
• Accuracy: Measures the proportion of correct predictions made by the model on the test dataset.
• Classwise Accuracy: Evaluates how well the model predicts each digit class (2 and 4).

💡 Why Naive Bayes Works Well for OCR

Naive Bayes is particularly suited for OCR tasks due to its ability to handle:

• Feature Independence: Each pixel is considered independently, making it computationally efficient for high-dimensional data.
• Small Training Data: Even with limited training data, Naive Bayes performs effectively, which is useful for OCR where labeled data might be sparse.

Limitations of Naive Bayes in OCR:

While Naive Bayes excels in simple cases, it may struggle with more complex features or where the relationships between pixels are not independent (e.g., continuous data such as house prices).

🚧 Future Improvements

• Advanced Techniques: Experiment with deep learning models such as Convolutional Neural Networks (CNNs) for more accurate and scalable OCR solutions.
• Hyperparameter Tuning: Enhance the model’s performance by tuning hyperparameters and experimenting with different machine learning algorithms.
• Real-Time OCR: Develop a real-time OCR system for applications like document scanning and live text recognition.

📝 Notebook Overview

In this notebook, we implement the core logic of the Naive Bayes OCR System. The steps followed are:

1. Preprocessing the Image: Convert raw pixel data into arrays and images for model input.
2. Feature Extraction: Extract pixel probabilities for the Naive Bayes model.
3. Training the Model: Train the Naive Bayes classifier on labeled images of handwritten digits.
4. Model Evaluation: Calculate accuracy, confusion matrix, and class-wise accuracy for evaluating model performance.

Here’s a snapshot of the OCR system in action:

🚀 Conclusion

This OCR system serves as an excellent starting point for recognizing handwritten digits using a Naive Bayes classifier. It demonstrates the efficiency of Naive Bayes in solving image classification tasks and lays the foundation for future work in more advanced OCR systems.

Feel free to fork this repository, contribute, and make enhancements! Let's build better OCR solutions together. 🤝✨

📜 License

This project is licensed under the MIT License. See the LICENSE file for more details.