Image Classification System

Detailed Explanation of the Code

1. Import Libraries:

   • Import necessary libraries including TensorFlow, Keras layers and models, ImageDataGenerator for data augmentation, Matplotlib for plotting, and NumPy for numerical operations.

2. Load CIFAR-10 Dataset:

   • Load the CIFAR-10 dataset which consists of 60,000 32x32 color images in 10 classes, with 6,000 images per class. Split into 50,000 training and 10,000 test images.

3. Normalize Images:

   • Normalize pixel values of the images to be between 0 and 1 for better convergence during training. This is done by dividing the pixel values by 255.

4. Split Training Data:

   • Use train_test_split from scikit-learn to split the training data into training and validation sets. Set aside 20% of the training data for validation purposes. This helps in monitoring the model's performance on unseen data during training.

5. Build the CNN Model:

   • Initialize a Sequential model and add the following layers:

     • First Convolutional Layer: Add a Conv2D layer with 32 filters, a kernel size of 3x3, ReLU activation function, and input shape of (32, 32, 3). Follow this with a MaxPooling2D layer with a pool size of 2x2.

     • Second Convolutional Layer: Add a Conv2D layer with 64 filters and a kernel size of 3x3, ReLU activation function. Follow this with a MaxPooling2D layer with a pool size of 2x2.

     • Third Convolutional Layer: Add a Conv2D layer with 64 filters and a kernel size of 3x3, ReLU activation function. Follow this with a MaxPooling2D layer with a pool size of 2x2.

     • Flatten Layer: Flatten the output from the convolutional layers to feed it into the dense layers.

     • Dense Layer: Add a Dense layer with 64 units and ReLU activation function.

     • Output Layer: Add a Dense layer with 10 units (one for each CIFAR-10 class) and softmax activation function to output probabilities for each class.

6. Compile the Model:

   • Compile the model using Adam optimizer, sparse categorical cross-entropy loss (suitable for integer labels), and accuracy as the evaluation metric.

7. Show Model Summary:

   • Display the model architecture with model.summary() to visualize the layer structure and parameter counts.

8. Data Augmentation:

   • Create an ImageDataGenerator instance to perform data augmentation. This includes random rotations, shifts, shear, zoom, and horizontal flips to artificially expand the training dataset and improve model generalization.

9. Train the Model:

   • Fit the model using the augmented data generated by ImageDataGenerator. Train for 30 epochs, with steps per epoch calculated as the number of training samples divided by the batch size. Validate on the validation dataset created earlier.

10. Evaluate the Model:

    • Evaluate the trained model on the test dataset to calculate the test loss and accuracy. Print the test accuracy.

11. Plot Training and Validation Accuracy/Loss:

    • Plot the training and validation accuracy and loss over epochs using Matplotlib to visualize the model's performance and check for overfitting or underfitting.

12. Save the Trained Model:

    • Save the trained model to a file named cnn_image_classification_model.h5 for later use.

13. Load the Saved Model:

    • Load the saved model from the file to make predictions on new data.

14. Preprocess and Predict New Images:

    • Define functions to preprocess new images (resize to 32x32, convert to array, and normalize) and to predict the class of a new image using the trained model.

15. Example Predictions:

    • Uncomment and modify the lines provided to test the prediction function on new images by specifying the image paths. The predicted class is printed for each image.

This step-by-step explanation covers the entire process from data loading and preprocessing, model building and training, to saving, loading, and using the trained model for predictions.