README.md

---
title: Digit Recognition with CNN
emoji: 🔢
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 4.19.2
app_file: app.py
pinned: false
---

# Digit Recognition Model

This model is trained to recognize handwritten digits from the MNIST dataset.

## Model Description

- **Model Type:** CNN with Attention
- **Task:** Image Classification
- **Input:** 28x28 grayscale images
- **Output:** Digit classification (0-9)

## Training

The model was trained on the MNIST dataset using a CNN architecture with attention mechanisms.

## Usage

```python
import tensorflow as tf
import numpy as np

# Load the model
model = tf.saved_model.load('https://huggingface.co/nivashuggingface/digit-recognition/resolve/main/saved_model')

# Prepare input
image = tf.keras.preprocessing.image.load_img("digit.png", target_size=(28, 28))
image = tf.keras.preprocessing.image.img_to_array(image)
image = image.astype('float32') / 255.0
image = np.expand_dims(image, axis=0)

# Make prediction
predictions = model(image)
predicted_digit = tf.argmax(predictions, axis=1).numpy()[0]
```

# AI Model Training Project

This project demonstrates a complete machine learning workflow from data preparation to model deployment, using the MNIST dataset with an innovative approach to digit recognition.

## Project Structure

```
.
├── data/                  # Dataset storage
├── models/               # Saved model files
├── src/                  # Source code
│   ├── data_preparation.py
│   ├── model.py
│   ├── training.py
│   ├── evaluation.py
│   └── deployment.py
├── notebooks/           # Jupyter notebooks for exploration
├── requirements.txt     # Project dependencies
└── README.md           # Project documentation
```

## Setup Instructions

1. Create a virtual environment:
```bash
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
```

2. Install dependencies:
```bash
pip install -r requirements.txt
```

3. Run the training pipeline:
```bash
python src/training.py
```

## Project Features

- Custom CNN architecture for robust digit recognition
- Data augmentation techniques
- Model evaluation and hyperparameter tuning
- Model deployment pipeline
- Performance monitoring

## Learning Concepts Covered

1. Data Preprocessing
   - Data loading and cleaning
   - Feature engineering
   - Data augmentation

2. Model Architecture
   - Custom CNN design
   - Layer configuration
   - Activation functions

3. Training Process
   - Loss functions
   - Optimizers
   - Learning rate scheduling
   - Early stopping

4. Evaluation
   - Metrics calculation
   - Cross-validation
   - Model comparison

5. Deployment
   - Model saving
   - Inference pipeline
   - Performance monitoring