Fine-tuning model_based on keras framework

Fine-tuning means to 'unfreeze' the top layers and train the unfrozen layers together with the newly added parts.

1. Data preparation

from keras.applications import VGG16
from keras import models
from keras import layers 
from keras import optimizers
import os,shutil
from keras.preprocessing.image import ImageDataGenerator
base_dir = "/kaggle/input/cats-and-dogs-small"
train_dir = os.path.join(base_dir,'train')
validation_dir = os.path.join(base_dir,'validation')

2. Model loading and freezing

conv_base = VGG16(weights='imagenet',include_top=False,input_shape=(150,150,3))
conv_base.trainable = True
set_trainable = False
for layer in conv_base.layers:
    if layer.name == 'block5_conv1':
        set_trainable = True
    if set_trainable:
        layer.trainable = True
    else:
        layer.trainable = False

3. Add a classifier

model = models.Sequential()
model.add(conv_base)
model.add(layers.Flatten())
model.add(layers.Dense(256,activation='relu'))
model.add(layers.Dense(1,activation='sigmoid'))

Check the parameters that need to be updated
Parameter general class

4. Generate iterator

train_datagen = ImageDataGenerator(rescale=1.0/255,rotation_range=40,width_shift_range=0.2,height_shift_range=0.2,shear_range=0.2,zoom_range=0.2,horizontal_flip=True,fill_mode='nearest')
test_datagen = ImageDataGenerator(rescale=1.0/255)
train_generator = train_datagen.flow_from_directory(train_dir,target_size=(150,150),batch_size=20,class_mode='binary')
validation_generator = test_datagen.flow_from_directory(validation_dir,target_size=(150,150),batch_size=20,class_mode='binary')

5. Generate the optimizer and start training

model.compile(loss='binary_crossentropy',optimizer=optimizers.RMSprop(lr=2e-5),metrics=['acc'])
history=model.fit_generator(train_generator,steps_per_epoch=100,epochs=50,validation_data=validation_generator,validation_steps=50)

6. Draw a curve graph

import matplotlib.pyplot as plt
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs = range(1,len(acc)+1)
plt.subplot(221)
plt.plot(epochs,acc,'bo',label='training acc')
plt.plot(epochs,val_acc,'b',label='validation acc')
plt.title('training and validation acc')
plt.legend()
plt.subplot(222)
plt.plot(epochs,loss,'bo',label='training loss')
plt.plot(epochs,val_loss,'b',label='validation loss')
plt.title('training and validation loss')
plt.legend()
plt.show()

Insert image description here

7. Smooth curve

import matplotlib.pyplot as plt
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']
def smooth_curve(points,factor=0.8):
    smooth_points = []
    for point in points:
        if smooth_points:
            previous = smooth_points[-1]
            smooth_points.append(previous*factor+point*(1-factor))
        else:
            smooth_points.append(point)
    return smooth_points
plt.subplot(221)
plt.plot(epochs,smooth_curve(acc),'bo',label='smoothed training acc')
plt.plot(epochs,smooth_curve(val_acc),'b',label='smoothed vaildation acc')
plt.title('training and validation acc')
plt.legend()
plt.subplot(222)
plt.plot(epochs,smooth_curve(loss),'bo',label='smoothed training loss')
plt.plot(epochs,smooth_curve(val_loss),'b',label='smoothed validation loss')
plt.title('training and validation loss')
plt.legend()
plt.show()