迁移学习精度差

Transfer learning bad accuracy

我有一项任务是根据缺陷对种子进行分类。我在 7 类 中有大约 14k 张图像(它们大小不等,有些 类 有更多照片,有些则更少)。我尝试从头开始训练 Inception V3,准确率达到了 90% 左右。然后我尝试使用带有 ImageNet 权重的预训练模型进行迁移学习。我从 applications 导入了 inception_v3 而没有顶部 fc 层,然后在文档中添加了我自己的 like。我以以下代码结束:

# Setting dimensions
img_width = 454
img_height = 227

###########################
# PART 1 - Creating Model #
###########################

# Creating InceptionV3 model without Fully-Connected layers
base_model = InceptionV3(weights='imagenet', include_top=False, input_shape = (img_height, img_width, 3))

# Adding layers which will be fine-tunned
x = base_model.output
x = GlobalAveragePooling2D()(x)
x = Dense(1024, activation='relu')(x)
predictions = Dense(7, activation='softmax')(x)

# Creating final model
model = Model(inputs=base_model.input, outputs=predictions)

# Plotting model
plot_model(model, to_file='inceptionV3.png')

# Freezing Convolutional layers
for layer in base_model.layers:
    layer.trainable = False

# Summarizing layers
print(model.summary())

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

##############################################
# PART 2 - Images Preproccessing and Fitting #
##############################################

# Fitting the CNN to the images

train_datagen = ImageDataGenerator(rescale = 1./255,
                                   rotation_range=30,
                                   width_shift_range=0.2,
                                   height_shift_range=0.2,
                                   shear_range = 0.2,
                                   zoom_range = 0.2,
                                   horizontal_flip = True,
                                   preprocessing_function=preprocess_input,)

valid_datagen = ImageDataGenerator(rescale = 1./255,
                                   preprocessing_function=preprocess_input,)

train_generator = train_datagen.flow_from_directory("dataset/training_set",
                                                    target_size=(img_height, img_width),
                                                    batch_size = 4,
                                                    class_mode = "categorical",
                                                    shuffle = True,
                                                    seed = 42)

valid_generator = valid_datagen.flow_from_directory("dataset/validation_set",
                                                    target_size=(img_height, img_width),
                                                    batch_size = 4,
                                                    class_mode = "categorical",
                                                    shuffle = True,
                                                    seed = 42)

STEP_SIZE_TRAIN = train_generator.n//train_generator.batch_size
STEP_SIZE_VALID = valid_generator.n//valid_generator.batch_size

# Save the model according to the conditions  
checkpoint = ModelCheckpoint("inception_v3_1.h5", monitor='val_acc', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)
early = EarlyStopping(monitor='val_acc', min_delta=0, patience=10, verbose=1, mode='auto')

#Training the model
history = model.fit_generator(generator=train_generator,
                         steps_per_epoch=STEP_SIZE_TRAIN,
                         validation_data=valid_generator,
                         validation_steps=STEP_SIZE_VALID,
                         epochs=25,
                         callbacks = [checkpoint, early])

但我得到了糟糕的结果:45% 的准确率。我认为它应该更好。我有一些假设可能会出错:

还是我做错了什么?

编辑:我post训练历史图。也许它包含有价值的信息:

EDIT2: 改变 InceptionV3 的参数:

对比VGG16:

如果您想使用 Keras 的 preprocess_input 方法预处理输入,请删除 rescale=1./255 参数。否则,保留 rescale 参数并删除 preprocessing_function 参数。另外,如果损失没有减少,请尝试较低的学习率,如 1e-4 或 3e-5 或 1e-5(Adam 优化器的默认学习率是 1e-3):

from keras.optimizers import Adam

model.compile(optimizer = Adam(lr=learning_rate), ...)

编辑: 添加训练图后,可以看到是不是在训练集上过拟合了。您可以:

  • 添加某种正则化,例如 Dropout 层,
  • 或通过减少最后一层之前的密集层中的单元数来减小网络大小。

@今天,我发现了一个问题。这是因为 Batch Normalization 层及其在冻结它们时的行为发生了一些变化。 Chollet 先生给出了一个解决方法,但我使用了 datumbox 制作的 Keras fork,这解决了我的问题。主要问题描述如下:

https://github.com/keras-team/keras/pull/9965

现在我的准确率约为 85%,我正在努力提高它。