Keras Convolution2D Input:检查模型输入时出错:期望的convolution2d_input_1

Question

我正在学习这个使用Keras创建图像分类器的教程。一旦我训练了模型，我将它保存到一个文件中，然后在测试脚本中将其重新加载到模型中，如下所示。

当我使用一个新的，从未见过的图像评估模型时，我得到以下异常:

错误:

Traceback (most recent call last):
  File "test_classifier.py", line 48, in <module>
    score = model.evaluate(x, y, batch_size=16)
  File "/Library/Python/2.7/site-packages/keras/models.py", line 655, in evaluate
    sample_weight=sample_weight)
  File "/Library/Python/2.7/site-packages/keras/engine/training.py", line 1131, in evaluate
    batch_size=batch_size)
  File "/Library/Python/2.7/site-packages/keras/engine/training.py", line 959, in _standardize_user_data
exception_prefix='model input')
  File "/Library/Python/2.7/site-packages/keras/engine/training.py", line 108, in standardize_input_data
str(array.shape))
Exception: Error when checking model input: expected convolution2d_input_1 to have shape (None, 3, 150, 150) but got array with shape (1, 3, 150, 198)`

是我所训练的模型的问题还是我如何调用评估方法的问题?

代码:

    from keras.preprocessing.image import ImageDataGenerator
    from keras.models import Sequential
    from keras.layers import Convolution2D, MaxPooling2D
    from keras.layers import Activation, Dropout, Flatten, Dense
    from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img
    import numpy as np
    img_width, img_height = 150, 150
    train_data_dir = 'data/train'
    validation_data_dir = 'data/validation'
    nb_train_samples = 2000
    nb_validation_samples = 800
    nb_epoch = 5
    model = Sequential()
    model.add(Convolution2D(32, 3, 3, input_shape=(3, img_width, img_height)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Convolution2D(32, 3, 3))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Convolution2D(64, 3, 3))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Flatten())
    model.add(Dense(64))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(1))
    model.add(Activation('sigmoid'))
    model.compile(loss='binary_crossentropy',
          optimizer='rmsprop',
          metrics=['accuracy'])
    model.load_weights('first_try.h5')
    img = load_img('data/test2/ferrari.jpeg')
    x = img_to_array(img)  # this is a Numpy array with shape (3, 150, 150)
    x = x.reshape( (1,) + x.shape )  # this is a Numpy array with shape (1, 3, 150, 150)
    y = np.array([0])
    score = model.evaluate(x, y, batch_size=16)`

Answer 1

问题是由于测试图像的大小错误。对我来说,

train_datagen.flow_from_directory(
        'C:\Users\...\train',  # this is the target directory
        target_size=(150, 150),  # all images will be resized to 150x150
        batch_size=32,
        class_mode='binary')

不能正常工作。所以我使用matlab命令来调整所有测试图像的大小，它工作得很好

Answer 2

问题有两个方面:

1. 测试图像大小错误。

2. 我必须将致密层从model.add(Dense(10))改为model.add(Dense(1))。

我还不明白如何让模型给我预测，但至少现在，模型评估运行。

Answer 3

我有同样的问题，并使用这个函数:目标文件夹(.jpg和.png)中的所有图像将被调整为高度和宽度。然后除以255。加上增加了一个尺寸(所需的输入形状)。

from scipy import misc
import os
def readImagesAsNumpyArrays(targetPath, i_height, i_width):
    files = os.listdir(targetPath)
    npList = list()
    for file in files:
        if ".jpg" or ".png" in str(file):
            path = os.path.join(targetPath, file)
            img = misc.imread(path)
            img = misc.imresize(img, (i_height, i_width))
            img = img * (1. / 255)
            img = img[None, :, :,: ]
            npList.append(img)
    return npList