🚀 EigenVision

EigenVision: Unlocking the power of computer vision with Python and Jupyter Notebooks.

---

🖼️ Project Overview

EigenVision is a practical demonstration of digit recognition using the MNIST dataset, Principal Component Analysis (PCA), and K-Nearest Neighbors (KNN) in Python.

---

📊 Project Overview

This project walks through the process of classifying handwritten digits from the MNIST dataset using:

• Data loading and visualization
• Data preprocessing (standardization)
• Dimensionality reduction with PCA
• Classification with K-Nearest Neighbors (KNN)
• Model evaluation and visualization of results

All code and analysis are contained in the Jupyter notebook: EigenVision.ipynb.

---

📁 Data

• File Used: train.csv (downloaded manually and loaded from local path)
• Shape: 42,000 samples × 785 columns (1 label + 784 pixel values)

---

🛠️ Workflow

1. Data Loading:

   • The MNIST training data is loaded from a local CSV file.
   • Data is split into features (X) and labels (y).

2. Visualization:

   • Individual digit images are visualized using matplotlib.

3. Train-Test Split:

   • Data is split into training and test sets (80/20 split).

4. KNN Baseline:

   • A KNN classifier is trained on the raw pixel data.
   • Baseline accuracy is computed.

5. Standardization:

   • Features are standardized using StandardScaler.

6. PCA:

   • Dimensionality is reduced using PCA (first to 200, then to 2 and 3 components for visualization).
   • Explained variance and eigenvectors are analyzed.

7. KNN on PCA Data:

   • KNN is retrained on PCA-transformed data.
   • Accuracy is evaluated again.

8. Visualization:

   • 2D and 3D scatter plots of the PCA-transformed data are created using Plotly.
   • Cumulative explained variance is plotted.

---

🏆 Results

• KNN on raw data: ~96.5% accuracy
• KNN on PCA (200 components): ~95.1% accuracy
• PCA Visualization:
  • 2D and 3D scatter plots show digit separation in reduced space.
  • Cumulative explained variance plot demonstrates how many components are needed to capture most variance.

---

📦 Requirements

• Python 3.13+
• pandas
• numpy
• scikit-learn
• matplotlib
• plotly

Install requirements with:

pip install pandas numpy scikit-learn matplotlib plotly

---

📂 Files

• EigenVision.ipynb — Main notebook with all code, plots, and results
• train.csv — MNIST training data (not included, download from Kaggle)

---

🚀 How to Run

1. Download train.csv from Kaggle and place it in the appropriate path (update the path in the notebook if needed).
2. Open EigenVision.ipynb in Jupyter Notebook or JupyterLab.
3. Run all cells to reproduce the analysis and results.

---

📄 License

This project is licensed under the MIT License.

---

📬 Contact

Project Maintainer: Aditya Sinha