การพัฒนาโมเดล Machine Learning : การจำแนกประเภทภาพ Deep Learning (CNN) โดยใช้ Python และ TensorFlow

import tensorflow as tf
import tensorflow_datasets as tfds
import matplotlib.pyplot as plt

(train_data, test_data), info = tfds.load(
    'cats_vs_dogs',
    split=['train[:80%]', 'train[80%:]'],
    with_info=True,
    as_supervised=True
)

print("Number of training images:", int(info.splits['train'].num_examples * 0.8))
print("Number of test images:", int(info.splits['train'].num_examples * 0.2))

IMG_SIZE = 128
BATCH_SIZE = 32

# ใช้ Data Augmentation เฉพาะกับชุด train
data_augmentation = tf.keras.Sequential([
    tf.keras.layers.RandomFlip('horizontal'),
    tf.keras.layers.RandomRotation(0.1),
    tf.keras.layers.RandomZoom(0.1)
])

def preprocess(image, label, augment=False):
    image = tf.image.resize(image, (IMG_SIZE, IMG_SIZE))
    image = image / 255.0
    if augment:
        image = data_augmentation(image)
    return image, label

# Apply preprocessing
train_data = train_data.map(lambda x, y: preprocess(x, y, augment=True), num_parallel_calls=tf.data.AUTOTUNE)
test_data = test_data.map(lambda x, y: preprocess(x, y), num_parallel_calls=tf.data.AUTOTUNE)

# ทำ batching และ prefetch
train_data = train_data.shuffle(1000).batch(BATCH_SIZE).prefetch(tf.data.AUTOTUNE)
test_data = test_data.batch(BATCH_SIZE).prefetch(tf.data.AUTOTUNE)

model = tf.keras.Sequential([
    tf.keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(IMG_SIZE, IMG_SIZE, 3)),
    tf.keras.layers.MaxPooling2D(2,2),

    tf.keras.layers.Conv2D(64, (3,3), activation='relu'),
    tf.keras.layers.MaxPooling2D(2,2),

    tf.keras.layers.Conv2D(128, (3,3), activation='relu'),
    tf.keras.layers.MaxPooling2D(2,2),

    tf.keras.layers.Dropout(0.3),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(128, activation='relu'),
    tf.keras.layers.Dropout(0.3),
    tf.keras.layers.Dense(1, activation='sigmoid')  
])

# คอมไพล์โมเดล
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])

# เพิ่ม early stopping
early_stop = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=3, restore_best_weights=True)

history = model.fit(
    train_data,
    validation_data=test_data,
    epochs=5,
    callbacks=[early_stop]
)

for images, labels in test_data.shuffle(100).take(1):
    for i in range(2):
        img = images[i]
        true_label = labels[i].numpy()
        prediction = model.predict(tf.expand_dims(img, axis=0))[0][0]
        predicted_label = 1 if prediction > 0.5 else 0

        print("Prediction:", "Dog" if predicted_label == 1 else "Cat")
        print("Actual label:", "Dog" if true_label == 1 else "Cat")
        print("---")

        plt.imshow(img)
        plt.axis("off")
        plt.show()

import tensorflow as tf
import matplotlib.pyplot as plt
from tensorflow.keras.callbacks import EarlyStopping

from google.colab import drive
drive.mount('/content/drive')

!ls "/content/drive/MyDrive/fruit_images"

# โหลด dataset จากโฟลเดอร์
train_data = tf.keras.utils.image_dataset_from_directory(
    '/content/drive/MyDrive/fruit_images',
    validation_split=0.2,
    subset="training",
    seed=123,
    image_size=(IMG_SIZE, IMG_SIZE),
    batch_size=BATCH_SIZE
)

test_data = tf.keras.utils.image_dataset_from_directory(
    '/content/drive/MyDrive/fruit_images',
    validation_split=0.2,
    subset="validation",
    seed=123,
    image_size=(IMG_SIZE, IMG_SIZE),
    batch_size=BATCH_SIZE
)

# Prefetch ให้เร็วขึ้น
train_data = train_data.prefetch(tf.data.AUTOTUNE)
test_data = test_data.prefetch(tf.data.AUTOTUNE)

model = tf.keras.Sequential([
    tf.keras.layers.Conv2D(32, (3,3), activation='relu', input_shape=(IMG_SIZE, IMG_SIZE, 3)),
    tf.keras.layers.MaxPooling2D(2,2),

    tf.keras.layers.Conv2D(64, (3,3), activation='relu'),
    tf.keras.layers.MaxPooling2D(2,2),

    tf.keras.layers.Conv2D(128, (3,3), activation='relu'),
    tf.keras.layers.MaxPooling2D(2,2),

    tf.keras.layers.Dropout(0.3),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(128, activation='relu'),
    tf.keras.layers.Dropout(0.3),
    tf.keras.layers.Dense(1, activation='sigmoid')
])

# คอมไพล์โมเดล
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])

# เพิ่ม EarlyStopping
early_stop = tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=3, restore_best_weights=True)

# เทรนโมเดล
history = model.fit(
    train_data,
    validation_data=test_data,
    epochs= 5,
    callbacks=[early_stop]
)

# ทดสอบโมเดลกับภาพตัวอย่าง
for images, labels in test_data.shuffle(100).take(1):
    for i in range(2):
        img = images[i]
        true_label = labels[i].numpy()
        prediction = model.predict(tf.expand_dims(img, axis=0))[0][0]
        predicted_label = 1 if prediction > 0.5 else 0

        print("Prediction:", "Grape" if predicted_label == 1 else "Apple")
        print("Actual label:", "Grape" if true_label == 1 else "Apple")
        print("---")

        plt.imshow(img / 255.0)
        plt.axis("off")
        plt.show()