Keras 迁移学习 - 损失不减少

Question

我正在尝试对 Cats and Dogs 数据集进行迁移学习，但问题是我已经阅读了许多文档，但我的损失并没有减少......有人可以帮忙吗？有一些根本性的错误。

我的损失保持在 50%，这是随机猜测。该模型根本就不是训练。我什至尝试更改 base_model.trainiable = True，仍然没有。

import os
for dirname, _, filenames in os.walk('/kaggle/input'):
    for filename in filenames:
        print(os.path.join(dirname, filename))
import random
import os
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

import tensorflow as tf
from tensorflow import keras
import zipfile
import shutil

def extract_files(source, target):
    unzip_source = zipfile.ZipFile(source, 'r')
    unzip_source.extractall(target)
    unzip_source.close()

os.listdir('/kaggle/input/')

extract_files('/kaggle/input/dogs-vs-cats/test1.zip', '/kaggle/working/')
extract_files('/kaggle/input/dogs-vs-cats/train.zip', '/kaggle/working/')

base_dir = ('/kaggle/working/')
base_dir

os.listdir('/kaggle/working/train')[:10]

print(len(os.listdir('/kaggle/working/train')))

os.makedirs('/kaggle/working/training/dogs')

os.makedirs('/kaggle/working/training/cats')

target_path = '/kaggle/working/training/'

os.listdir(target_path)

def copy_file(file_path, file_name, target):
    source_path = os.path.join(file_path, file_name)
    final_path = os.path.join(target_path, target)
    shutil.copy(source_path, final_path)


def separate_files(file_path):
    for img in os.listdir(file_path):
        
        if img.split('.')[0] == 'cat':
            copy_file(file_path, img, os.path.join(target_path, 'cats'))
        else:
            copy_file(file_path, img, os.path.join(target_path, 'dogs'))
    
    
        

separate_files(os.path.join(base_dir, 'train'))

os.listdir(os.path.join(base_dir,'training','cats'))[:20]

print(len(os.listdir(os.path.join(base_dir,'training','cats'))))
print(len(os.listdir(os.path.join(base_dir,'training','dogs'))))

cat_folder = os.path.join(os.path.join(base_dir,'training','cats'))
dog_folder = os.path.join(os.path.join(base_dir,'training','dogs'))

data_dir = os.path.join(base_dir, 'training')

plt.figure(figsize=(10,10))

for i in range(15):
    sub_dir_name = np.random.choice(os.listdir(data_dir))
    sub_dir = os.path.join(data_dir, sub_dir_name)
    img_file = os.path.join(sub_dir, np.random.choice([x for x in os.listdir(sub_dir)] ))
    
    plt.subplot(5,5,i+1)
    plt.xticks([])
    plt.yticks([])
    plt.grid(False)
    img = plt.imread(img_file)
    plt.imshow(img)
    plt.xlabel(f'dir: {sub_dir_name}')
    
plt.show()


from tensorflow.keras.preprocessing.image import ImageDataGenerator

train_gen = ImageDataGenerator(rescale = 1.0/255.0, 
                              shear_range= 0.2,
                              zoom_range= 0.3,
                              width_shift_range = 0.2,
                              height_shift_range = 0.2,
                              horizontal_flip= True,
                              vertical_flip= True,
                              rotation_range= 45,
                              validation_split= 0.2,
                              )

valid_gen = ImageDataGenerator(rescale = 1.0/255.0,
                              shear_range= 0.2,
                              zoom_range= 0.3,
                              horizontal_flip= True,
                              vertical_flip= True,
                              rotation_range= 45,
                              validation_split= 0.2,
                              )
                               
train_ds = train_gen.flow_from_directory(data_dir,
                                         target_size= (160,160),
                                         shuffle= True,
                                         batch_size = 128,
                                         seed= 7,
                                         subset= 'training'
                                         )    


valid_ds = valid_gen.flow_from_directory(data_dir,
                                         target_size= (160,160),
                                         shuffle= True,
                                         batch_size = 128,
                                         seed= 7,
                                         subset= 'validation'
                                         )   

# from keras.applications.inception_resnet_v2 import InceptionResNetV2
# from keras.applications.inception_resnet_v2 import preprocess_input
# from keras.applications.inception_resnet_v2 import decode_predictions
# from sklearn.metrics import accuracy_score


tf.keras.backend.clear_session()


base_model = tf.keras.applications.MobileNetV2(input_shape = (160,160,3), 
                                               weights= 'imagenet', 
                                               include_top = False)

base_model.trainable = False

base_model.summary()

global_avg_layer = tf.keras.layers.GlobalAveragePooling2D()


model = tf.keras.Sequential([base_model,
                            tf.keras.layers.Flatten(),
                            tf.keras.layers.Dense(512, activation = 'relu'),
                            tf.keras.layers.Dropout(0.5), 
                            tf.keras.layers.Dense(1, activation = 'sigmoid')])

model.summary()

model.compile(optimizer= tf.keras.optimizers.Adam(learning_rate= 0.0001), 
             loss = tf.keras.losses.BinaryCrossentropy(from_logits= True),
             metrics = ['accuracy'])

early_stop = tf.keras.callbacks.EarlyStopping(patience = 10)
reduce_on_plateau = tf.keras.callbacks.ReduceLROnPlateau(patience = 5)

history = model.fit(train_ds, epochs=200, validation_data= valid_ds, callbacks= [early_stop, reduce_on_plateau])