Keras 入门教程：2025 新版预训练模型库实战图像识别全流程

pip install tensorflow
pip install keras

import keras
print(keras.__version__)

from tensorflow.keras.datasets import cifar10

(x_train, y_train), (x_test, y_test) = cifar10.load_data()

x_train = x_train.astype('float32') / 255.0
x_test = x_test.astype('float32') / 255.0

from tensorflow.keras.utils import to_categorical

y_train = to_categorical(y_train, )
y_test = to_categorical(y_test, )

from tensorflow.keras.applications import ResNet50

base_model = ResNet50(weights='imagenet', include_top=False, input_shape=(, , ))

for layer in base_model.layers:
    layer.trainable = False

from tensorflow.keras.layers import GlobalAveragePooling2D

x = GlobalAveragePooling2D()(base_model.output)

from tensorflow.keras.layers import Dense

x = Dense(, activation='relu')(x)
predictions = Dense(, activation='softmax')(x)

from tensorflow.keras.models import Model

model = Model(inputs=base_model.input, outputs=predictions)
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

from tensorflow.keras.preprocessing.image import ImageDataGenerator

datagen = ImageDataGenerator(
    rotation_range=,
    width_shift_range=0.2,
    height_shift_range=0.2,
    horizontal_flip=True
)
datagen.fit(x_train)

history = model.fit(
    datagen.flow(x_train, y_train, batch_size=),
    epochs=,
    validation_data=(x_test, y_test)
)

from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping

checkpoint = ModelCheckpoint(
    'best_model.h5',
    monitor='val_accuracy',
    save_best_only=True,
    mode='max'
)

early_stopping = EarlyStopping(
    monitor='val_loss',
    patience=,
    mode='min'
)

history = model.fit(
    datagen.flow(x_train, y_train, batch_size=),
    epochs=,
    validation_data=(x_test, y_test),
    callbacks=[checkpoint, early_stopping]
)

loss, accuracy = model.evaluate(x_test, y_test)
print(f'Test loss: {loss:.4f}')
print(f'Test accuracy: {accuracy:.4f}')

import numpy as np

# 假设我们有一张新的图像，尺寸为32x32x3
new_image = np.expand_dims(new_image, axis=)
new_image = new_image.astype('float32') / 255.0

predictions = model.predict(new_image)
predicted_class = np.argmax(predictions)

for layer in base_model.layers[-:]:
    layer.trainable = True

model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])

history = model.fit(
    datagen.flow(x_train, y_train, batch_size=),
    epochs=,
    validation_data=(x_test, y_test),
    callbacks=[checkpoint, early_stopping]
)

model.save('image_classification_model.h5')

converter = tf.lite.TFLiteConverter.from_keras_model(model)
tflite_model = converter.convert()
with open('image_classification_model.tflite', 'wb') as f:
    f.write(tflite_model)