问经过卷积神经网络后，4维数据变为2维数据

EN

from keras.utils.np_utils import to_categorical
import pandas as pd 
import numpy as np
import os
from keras.applications.vgg16 import VGG16
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential, Model
from keras.layers import Input, Activation, Dropout, Flatten, Dense,Concatenate, concatenate,Reshape, BatchNormalization, Merge
from keras.preprocessing.image import ImageDataGenerator
from keras import optimizers
from keras.optimizers import Adagrad
from sklearn import preprocessing
from scipy.misc import imread
import time
from PIL import Image 
import cv2
img_width, img_height = 60, 60
img_list = []
BR_list = []
label_list = []
data_num = 1630

folder1 = "cut2/train/sugi/"

folder2 = "cut2/train/hinoki/"

def imgConvert(file_path):
    img = imread(file_path,flatten = True)
    img = np.arange(1*3*60*60).reshape((60,60,3))
    img = np.array(img).reshape(60,60,3)
    img = img.astype("float32")
    return img

def B_and_R(img_path):
    img = cv2.imread(img_path)
    B = 0
    R = 0

    for i in range(25,35):
    #print(i)
        for j in range(25,35):
            B = B+img[i,j,0]
            R = R+img[i,j,2]
        #(j)
        #(img[i,j])

    ave_B = B/100
    ave_R = R/100

    BR = ave_B/ave_R
    return BR

def getData(path,pollen):
    for the_file in os.listdir(path):
        #print(the_file)
        file_path = os.path.join(path, the_file)
        B_over_R = B_and_R(file_path)
        img_arr = imgConvert(file_path)
        #writer.writerow([img_arr,B_over_R,"sugi"])
        img_list.append(img_arr)
        BR_list.append(B_over_R)
        lb = np.zeros(2)
        if pollen == "sugi":
            lb[0] +=1
        else:
            lb[1] +=1
        label_list.append(lb)

if __name__ == '__main__':

    getData(folder1,"sugi")
    getData(folder2,"hinoki")

    img_arr = np.array(img_list)
    print(img_arr.shape)
    #.reshape(img_list[0],1,img_width,img_height)
    img_arr.astype("float32")
    img_arr /= 255
    print(img_arr.shape)
    img_array = np.expand_dims(img_arr, axis = 0)
    img_array = img_array[0,:,:,:,:]
    print(img_array.shape)

    """
    datagen = ImageDataGenerator(
        featurewise_center=True,
        featurewise_std_normalization=True,
        rotation_range=20,
        width_shift_range=0.2,
        height_shift_range=0.2,
        horizontal_flip=True)
    datagen.fit(img_array)
    """
    #img_array = img_array.reshape(img_array[0],1,img_width,img_height)

    print(img_array.shape)
    label_arr = np.array(label_list)
    print(label_arr.shape)
    #label_array = np.expand_dims(label_arr, axis = 0)
    #label_array = label_array[0,:,:,:,:]

    BR_arr = np.array(BR_list)
    print(BR_arr.shape)
    #BR_array = np.expand_dims(BR_arr, axis = 0)
    #BR_array = BR_array[0,:,:,:,:]

    #print(len([img_arr,BR_arr]))

    input_tensor = Input(shape=(img_width, img_height,3))
    vgg16 = VGG16(include_top=False, weights='imagenet', input_tensor=input_tensor)

    # FC層の作成
    top_model = Sequential()
    top_model.add(Flatten(input_shape=vgg16.output_shape[1:]))

#print(top_model.summary())

    # VGG16とFC層を結合してモデルを作成
    branch1 = Model(input=vgg16.input, output=top_model(vgg16.output))
#model.summary()
    print(branch1.summary())


    branch2 = Sequential()
    branch2.add(Dense(1, input_shape=(data_num,1), activation='sigmoid'))
    #branch1.add(Reshape(BR.shape, input_shape = BR.shape))
    branch2.add(BatchNormalization())
    branch2.add(Flatten())

    print(branch2.summary())
    merged = Merge([branch1, branch2], mode = "concat")

    model = Sequential()
    model.add(merged)
    model.add(Dense(256, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(2, activation='softmax'))

   #last_model = Model(input = [branch1.input,branch2.input],output=model())
    print(model.summary())
    model.compile(loss='categorical_crossentropy',
              optimizer=optimizers.SGD(lr=1e-3, momentum=0.9),
              metrics=['accuracy'])
    print(img_array.shape)
    model.fit([img_array,BR_arr], label_arr,
            epochs=5, batch_size=100, verbose=1)