CNN Keras

#1. IMPORT FILE
import keras
from keras import Sequential
from keras.layers import Conv2D, Dense, Dropout, MaxPooling2D, Flatten
from keras.preprocessing import image
import numpy as np

#2. IMPORT MNIST DATASET
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()

# Expand dimensions for CNN (28x28x1)
x_train = np.expand_dims(x_train, axis=-1)
x_test  = np.expand_dims(x_test, axis=-1)

#3. NORMALIZATION
x_train = x_train.astype("float32") / 255.0
x_test  = x_test.astype("float32") / 255.0

# Convert to keras dataset batches
train_ds = keras.utils.image_dataset_from_directory  # dummy reference to match structure
train_ds = keras.utils.Sequence      # not used but kept to avoid structural change

# Manual batching
batch_size = 32
train_data = keras.utils.to_categorical(y_train, num_classes=10)
test_data  = keras.utils.to_categorical(y_test, num_classes=10)

#4. MODEL CREATION
model = Sequential([
    Conv2D(32, (5, 5), activation="relu", padding="same", input_shape=(28, 28, 1)),
    MaxPooling2D(padding="same"),
    Conv2D(64, (5, 5), activation="relu", padding="same"),
    MaxPooling2D(padding="same"),
    Flatten(),
    Dense(1024, activation="relu"),
    Dropout(0.2),
    Dense(10, activation="softmax")  # 10 for MNIST 0–9 digits
])

# OPTIONAL
# model.summary()

#5. MODEL COMPILE AND TRAINING
model.compile(
    optimizer="adam",
    loss="categorical_crossentropy",
    metrics=["accuracy"]
)

model.fit(
    x_train, train_data,
    epochs=5,
    batch_size=batch_size,
    validation_data=(x_test, test_data)
)

#6. IMAGE ARRAY (Predict using a test image)
img = x_test[0]
img_array = np.expand_dims(img, axis=0)

#7. MODEL PREDICTION
pred = model.predict(img_array)
pred_class = np.argmax(pred, axis=1)[0]

#8. LABEL PRINT
print("Predicted label:", pred_class)