recognizer_complex_cnn.py

import numpy as np
import pandas as pd
from tensorflow.keras import layers, models, regularizers
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.callbacks import ReduceLROnPlateau, EarlyStopping
from sklearn.model_selection import train_test_split

# Load the training and test data
train = pd.read_csv('train.csv')
test = pd.read_csv('test.csv')

# Separate labels and features
y_df = train[['label']]
x_df = train.drop(['label'], axis=1)

# Convert to numpy arrays
y = y_df.to_numpy().flatten()
X = x_df.to_numpy()

# Normalize the images to the range [0, 1]
X = X / 255.0

# Reshape the images to include the channel dimension
X = X.reshape(X.shape[0], 28, 28, 1)

# Split the data into training and validation sets
X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.1, random_state=42)

# Data augmentation
datagen = ImageDataGenerator(
    rotation_range=15,
    width_shift_range=0.15,
    height_shift_range=0.15,
    zoom_range=0.2,
    shear_range=0.15
)
datagen.fit(X_train)

# Modify the model with additional layers and regularization
model = models.Sequential([
    layers.Input(shape=(28, 28, 1)),
    layers.Conv2D(32, (3, 3), activation='relu', padding='same'),
    layers.BatchNormalization(),
    layers.Conv2D(32, (3, 3), activation='relu', padding='same'),
    layers.MaxPooling2D((2, 2)),
    layers.Dropout(0.3),

    layers.Conv2D(64, (3, 3), activation='relu', padding='same'),
    layers.BatchNormalization(),
    layers.Conv2D(64, (3, 3), activation='relu', padding='same'),
    layers.MaxPooling2D((2, 2)),
    layers.Dropout(0.3),

    layers.Conv2D(128, (3, 3), activation='relu', padding='same'),
    layers.BatchNormalization(),
    layers.Conv2D(128, (3, 3), activation='relu', padding='same'),
    layers.MaxPooling2D((2, 2)),
    layers.Dropout(0.4),

    layers.Flatten(),
    layers.Dense(256, activation='relu', kernel_regularizer=regularizers.l2(0.001)),
    layers.BatchNormalization(),
    layers.Dropout(0.5),
    layers.Dense(10, activation='softmax')
])

# Compile the model
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])

# Callbacks for learning rate adjustment and early stopping
callbacks = [
    ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=3, verbose=1),
    EarlyStopping(monitor='val_loss', patience=5, restore_best_weights=True)
]

# Train the model with callbacks
model.fit(datagen.flow(X_train, y_train, batch_size=64),
          epochs=50, 
          validation_data=(X_val, y_val),
          callbacks=callbacks)

# Save the model
model.save('digit_recognition_model.keras')

# Prepare the test data
test = test.to_numpy()
test = test / 255.0
test = test.reshape(test.shape[0], 28, 28, 1)

# Predict the labels for the test data
predictions = model.predict(test)
predicted_labels = np.argmax(predictions, axis=1)

# Save the predictions to a CSV file
submission = pd.DataFrame({
    'ImageId': range(1, len(predicted_labels) + 1),
    'Label': predicted_labels
})
submission.to_csv('submission.csv', index=False)