🔢 Project Contribution: Handwritten_Digit_Prediction_Program
I would like to contribute my Machine Learning + Web App project titled "Handwritten_Digit_Prediction_Program" under the Deep_Learning, Digit_Recognition, or Web_Integrated_AI section of this repository as part of GirlScript Summer of Code 2025 (GSSoC'25).
📌 Project Overview:
Handwritten_Digit_Prediction_Program is a deep learning-based project designed to classify handwritten digits (0–9) using a Multilayer Perceptron (MLP). The model is trained on the MNIST dataset, which contains thousands of handwritten digit images, and is capable of predicting digits in real time via a user-friendly web interface.
This project demonstrates the application of neural networks for image classification, even without convolutional layers, using only fully connected dense layers and pixel data.
🧠 Model Overview:
Framework: Python with TensorFlow/Keras, NumPy, and scikit-learn
Architecture: Multilayer Perceptron (Fully Connected Layers)
Input: 28x28 grayscale images flattened to 1D vectors
Preprocessing: Normalization, grayscale conversion, reshaping
Output: 10-class softmax (digits 0 to 9)
Optimization: Backpropagation with categorical crossentropy loss
Evaluation:
Accuracy, precision, recall
Confusion matrix for detailed error analysis
🧰 Features:
🔁 Trained models saved in .h5 and .keras formats
📁 MNIST data handled from .idx format
🧠 Clean preprocessing and training scripts
🌐 Flask-powered backend (app.py)
🖥️ HTML + CSS + JavaScript frontend for UI
🎯 Real-time image upload and prediction output
📈 Performance evaluated using multiple classification metrics
📦 Handwritten_Digit_Prediction_Program/
├── DATABASE/ # Raw MNIST dataset (IDX format)
├── model/ # Trained MLP models (.h5, .keras)
├── static/ # Frontend styling and scripts
│ ├── style.css
│ └── script.js
├── templates/
│ └── webpage.html # User input and prediction display
├── app.py # Flask app for backend prediction
├── .ipynb_checkpoints/ # Jupyter notebook backups
├── .idea/ # IDE-specific configs
└── README.md # (Can be improved for instructions)
✅ Deliverables:
✔️ Modular, well-documented Python code
✔️ Trained digit recognition model (MLP)
✔️ Flask web app for real-time predictions
✔️ Frontend interface to upload images and display results
✔️ Performance metrics including confusion matrix
✔️ Inference-ready .h5/.keras model files
👨💻 Full Name:
Shimanshu Chouhan
🎓 Participant Role:
GSSoC'25 Contributor
🔢 Project Contribution: Handwritten_Digit_Prediction_ProgramI would like to contribute my Machine Learning + Web App project titled "Handwritten_Digit_Prediction_Program" under the Deep_Learning, Digit_Recognition, or Web_Integrated_AI section of this repository as part of GirlScript Summer of Code 2025 (GSSoC'25).
📌 Project Overview:
Handwritten_Digit_Prediction_Program is a deep learning-based project designed to classify handwritten digits (0–9) using a Multilayer Perceptron (MLP). The model is trained on the MNIST dataset, which contains thousands of handwritten digit images, and is capable of predicting digits in real time via a user-friendly web interface.
This project demonstrates the application of neural networks for image classification, even without convolutional layers, using only fully connected dense layers and pixel data.
🧠 Model Overview:
Framework: Python with TensorFlow/Keras, NumPy, and scikit-learn
Architecture: Multilayer Perceptron (Fully Connected Layers)
Input: 28x28 grayscale images flattened to 1D vectors
Preprocessing: Normalization, grayscale conversion, reshaping
Output: 10-class softmax (digits 0 to 9)
Optimization: Backpropagation with categorical crossentropy loss
Evaluation:
Accuracy, precision, recall
Confusion matrix for detailed error analysis
🧰 Features:
🔁 Trained models saved in .h5 and .keras formats
📁 MNIST data handled from .idx format
🧠 Clean preprocessing and training scripts
🌐 Flask-powered backend (app.py)
🖥️ HTML + CSS + JavaScript frontend for UI
🎯 Real-time image upload and prediction output
📈 Performance evaluated using multiple classification metrics
📦 Handwritten_Digit_Prediction_Program/
├── DATABASE/ # Raw MNIST dataset (IDX format)
├── model/ # Trained MLP models (.h5, .keras)
├── static/ # Frontend styling and scripts
│ ├── style.css
│ └── script.js
├── templates/
│ └── webpage.html # User input and prediction display
├── app.py # Flask app for backend prediction
├── .ipynb_checkpoints/ # Jupyter notebook backups
├── .idea/ # IDE-specific configs
└── README.md # (Can be improved for instructions)
✅ Deliverables:
✔️ Modular, well-documented Python code
✔️ Trained digit recognition model (MLP)
✔️ Flask web app for real-time predictions
✔️ Frontend interface to upload images and display results
✔️ Performance metrics including confusion matrix
✔️ Inference-ready .h5/.keras model files
👨💻 Full Name:
Shimanshu Chouhan
🎓 Participant Role:
GSSoC'25 Contributor