https://github.com/ratna-babu/handwritten-digit-classification-using-cnn
https://github.com/ratna-babu/handwritten-digit-classification-using-cnn
Last synced: 7 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/ratna-babu/handwritten-digit-classification-using-cnn
- Owner: Ratna-Babu
- Created: 2025-01-21T16:36:05.000Z (9 months ago)
- Default Branch: main
- Last Pushed: 2025-01-21T17:17:57.000Z (9 months ago)
- Last Synced: 2025-01-21T17:36:59.510Z (9 months ago)
- Language: Jupyter Notebook
- Size: 80.1 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Handwritten-Digit-Classification-using-CNN
This project involves building a Convolutional Neural Network (CNN) to classify handwritten digits from the MNIST dataset. The MNIST dataset consists of 60,000 training images and 10,000 test images of handwritten digits ranging from 0 to 9. The goal is to accurately recognize these digits through the CNN model.
This project demonstrates how to build and train a Convolutional Neural Network (CNN) to recognize handwritten digits from the MNIST dataset using TensorFlow and Keras.
## Project Structure
The project consists of three main parts:
1. **Data Loading and Preprocessing:**
```python
import tensorflow as tf
# MNIST dataset
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
# Invert the colors of the images
x_train = 255 - x_train
x_test = 255 - x_test
# Preprocessing the Images
x_train = x_train.astype('float32') / 255
x_test = x_test.astype('float32') / 255
# Reshape the images to include the channel dimension
x_train = x_train.reshape(x_train.shape[0], 28, 28, 1)
x_test = x_test.reshape(x_test.shape[0], 28, 28, 1)
# One-hot encode the labels
y_train = tf.keras.utils.to_categorical(y_train, 10)
y_test = tf.keras.utils.to_categorical(y_test, 10)
```
- Imports the MNIST dataset using `tf.keras.datasets.mnist.load_data()`.
- Inverts the colors of the images for better model performance.
- Normalizes pixel values to the range [0, 1].
- Reshapes images to include the channel dimension.
- One-hot encodes the labels for multi-class classification.
2. **CNN Model Building and Training:**
- Builds a CNN model using `Sequential` with layers like `Conv2D`, `MaxPooling2D`, `Flatten`, and `Dense`.
```python
# CNN model
model = Sequential([
Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=(28, 28, 1)),
MaxPooling2D(pool_size=(2, 2)),
Conv2D(64, kernel_size=(3, 3), activation='relu'),
MaxPooling2D(pool_size=(2, 2)),
Flatten(),
Dense(128, activation='relu'),
Dense(10, activation='softmax')
])
```
- Compiles the model with the Adam optimizer, categorical cross-entropy loss, and accuracy metric.
```python
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
```
- Trains the model using `fit()` with specified epochs, batch size, and validation split.
```python
model.fit(x_train, y_train, epochs=10, batch_size=200, validation_split=0.2)
```
- Saves the trained model to a file named 'mnist_cnn.h5'.
```python
model.save('mnist_cnn.h5')
```
3. **Prediction:**
- Loads the trained model using `load_model()`.
- Defines a `predict_digit()` function to preprocess an input image and make a prediction.
- Defines a `display_prediction()` function to show the image with the prediction
- Calls `predict_digit()` with an image path to predict the digit.
- Prints the predicted digit to the console.
## Usage
1. **Install Dependencies:**
`pip install tensorflow==2.12.0`
`pip install keras==2.12.0`
`pip install Pillow==9.5.0`
`pip install numpy==1.24.3`
`pip install opencv-python==4.8.0`
`pip install matplotlib==3.7.2`
2. **Run the Code:**
Execute the code in a Google Colab environment or a Jupyter Notebook.
3. **Prediction:**
- Replace `'2.png'` in the `image_path` variable with the path to your image.
- Run the code to see the prediction result printed to the console.
## Model Performance
The model achieves high accuracy on the MNIST test set. To see the output, run the code.
## Ouput
## Note
- The model is trained on inverted images (black digits on a white background).
- Make sure to have the necessary libraries installed before running the code.