https://towardsdatascience.com/keras-transfer-learning-for-beginners-6c9b8b7143e
https://www.jianshu.com/p/763e48eaeebc
遷移學習：站在巨人的肩膀上，使用成熟的model(h5)，只需要替換最後幾層
好處：
1：不需要一個非常大的訓練數據集。
2：不需要太多的計算能力，因爲我們使用的是預先訓練好的權重，只需要學習最後幾層的權重。
成熟的model有：
- VGG (e.g. VGG16 or VGG19).
- GoogLeNet (e.g. InceptionV3).
- Residual Network (e.g. ResNet50).
層說明：
conv網絡前幾層的過濾器學習識別顏色和某些水平和垂直線。
接下來的幾層慢慢地學習如何使用在前幾層中學習到的線條和顏色來識別細小的形狀。
然後，下一層學習識別紋理，然後部分對象，如腿、眼睛、鼻子等。
最後，最後一層中的過濾器被諸如狗、汽車等整個物體激活。
例子1：直接使用模型，判斷出圖像裏是杜賓狗。（說明杜賓狗在既定的1000個分類中）
# example of using a pre-trained model as a classifier
from keras.preprocessing.image import load_img
from keras.preprocessing.image import img_to_array
from keras.applications.vgg16 import preprocess_input
from keras.applications.vgg16 import decode_predictions
from keras.applications.vgg16 import VGG16
# load an image from file
image = load_img('ai/dog.jpg', target_size=(224, 224))
# convert the image pixels to a numpy array
image = img_to_array(image)
# reshape data for the model
image = image.reshape((1, image.shape[0], image.shape[1], image.shape[2]))
# prepare the image for the VGG model
image = preprocess_input(image)
# load the model
model = VGG16()
# predict the probability across all output classes
yhat = model.predict(image)
# convert the probabilities to class labels
label = decode_predictions(yhat)
# retrieve the most likely result, e.g. highest probability
label = label[0][0]
# print the classification
print('%s (%.2f%%)' % (label[1], label[2]*100))
例子2，替換最後輸出層，重新定義輸出
import pandas as pd
import numpy as np
import os
import keras
import matplotlib.pyplot as plt
from keras.layers import Dense,GlobalAveragePooling2D
from keras.applications import MobileNet
from keras.preprocessing import image
from keras.applications.mobilenet import preprocess_input
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Model
from keras.optimizers import Adam
#imports the mobilenet model and discards the last 1000 neuron layer.
base_model=MobileNet(weights='imagenet',include_top=False)
# define output
x=base_model.output
x=GlobalAveragePooling2D()(x)
x=Dense(1024,activation='relu')(x) #we add dense layers so that the model can learn more complex functions and classify for better results.
x=Dense(1024,activation='relu')(x) #dense layer 2
x=Dense(512,activation='relu')(x) #dense layer 3
preds=Dense(3,activation='softmax')(x) #final layer with softmax activation
#now a model has been created based on our architecture
model=Model(inputs=base_model.input,outputs=preds)
# 前20層所有的權重設置爲不可訓練。我們將只訓練20層以後的權重
for layer in model.layers[:20]:
layer.trainable=False
for layer in model.layers[20:]:
layer.trainable=True
# use ImageDataGenerator to get traning data
train_datagen=ImageDataGenerator(preprocessing_function=preprocess_input) #included in our dependencies
train_generator=train_datagen.flow_from_directory('./train/', # this is where you specify the path to the main data folder
target_size=(224,224),
color_mode='rgb',
batch_size=32,
class_mode='categorical',
shuffle=True)
# compile
# Adam optimizer
# loss function will be categorical cross entropy
# evaluation metric will be accuracy
model.compile(optimizer='Adam',loss='categorical_crossentropy',metrics=['accuracy'])
step_size_train=train_generator.n//train_generator.batch_size
model.fit_generator(generator=train_generator,
steps_per_epoch=step_size_train,
epochs=5)
