MNIST Image Classification with TensorFlow

This project utilizes TensorFlow to build and train a model for classifying handwritten digits (0-9) from the MNIST dataset. Achieved a test accuracy of 98.5%.

Project Structure:

• dataset_exploration.py: Explores and visualizes the MNIST dataset.
• image_classification.py: Builds, trains, and evaluates a CNN for classifying MNIST digits.

dataset_exploration.py

This script performs the following tasks:

1. Loads the MNIST dataset using tensorflow_datasets with shuffling enabled to ensure random data access during training.
2. Explores dataset features: Extracts and saves information about the dataset features (image shape, data type, number of classes) to a text file named exploratory_output.txt.
3. Visualizes Sample Images: Displays and saves 10 randomly chosen images with their corresponding labels as a PNG image named mnist_sample_images.png.

image_classification.py

This script trains a Convolutional Neural Network (CNN) model for MNIST image classification:

1. Loads MNIST data: Loads the dataset using tensorflow.keras.datasets.mnist.
2. Preprocesses data:
   • Reshapes images to a format suitable for CNNs (28x28 pixels with one color channel).
   • Normalizes pixel values between 0 and 1 for better training performance.
   • One-hot encodes labels for multi-class classification (0 to 9).
3. Defines the CNN model:
   • Uses a sequential model with the following layers:
     • Conv2D: Applies a convolutional filter to extract features from the image.
     • MaxPooling2D: Reduces the spatial dimensionality of the data.
     • Flatten: Flattens the data into a single dimension for feeding to the dense layers.
     • Dense: Fully connected layer with ReLU activation for non-linearity.
     • Dense: Output layer with 10 units and softmax activation for probability distribution (one for each digit class).
4. Compiles the model:
   • Sets the optimizer ('adam') and loss function ('categorical_crossentropy') for training.
   • Defines accuracy as the evaluation metric.
5. Trains the model: Fits the model to the training data (x_train and y_train) for 10 epochs. Uses validation data (x_test and y_test) to monitor performance during training.
6. Evaluates the model: Evaluates the model's accuracy on the test set and prints the test accuracy. Achieved a test accuracy of 98.5%.

Running the Project:

1. Install all libraries found in requirements.txt (pip install -r requirements.txt).
2. Run python dataset_exploration.py to explore the data and save visualizations.
3. Run python image_classification.py to train and evaluate the model. The script will print the test accuracy.

Future Work: This is a basic implementation. Further improvements may include:

• Experimentation with different hyperparameters (learning rate, number of layers, etc.)
• Implementation of data augmentation techniques to improve model robustness.
• Exploration of more advanced CNN architectures (e.g., VGG16, ResNet).