Digit-Recognition-Using-ML

This project demonstrates how a robot can “see” and recognize handwritten digits using Machine Learning. We build a complete ML pipeline to classify small grayscale images (0–9) simulating what a robot might encounter when reading room signs, control panels, or numeric indicators.

It is designed as a beginner-friendly, end-to-end project, showcasing how to go from raw data to a working prediction system with clear, interpretable results.

Objectives

1. Simulate a robot’s digit perception task
2. Build and evaluate a classification model
3. Visualize predictions and performance

Dataset

We use the built-in load_digits dataset from scikit-learn.

Requirements

1. Python ≥ 3.7
2. Jupyter Notebook
3. numpy
4. pandas
5. matplotlib
6. seaborn
7. scikit-learn

Images: 1797 handwritten digit samples

Size: 8×8 pixels (grayscale)

Labels: 0–9

Each image is flattened into 64 pixel values to serve as input features.

🔷 Highlights

1. Data exploration & visualization
2. Random Forest classifier with ~98% accuracy
3. Confusion matrix & classification report
4. Predicted images vs. actual labels visualized

Results:

1. Test Accuracy: ~98%
2. Very low misclassification rate
3. Confident predictions with interpretable visualizations

Sample output:

Heatmap of confusion matrix
Example images annotated with predicted and true labels