如何在Keras上对MNIST数据集进行训练后，使用CNN预测自己的图像？

Question

我已经制作好了一个卷积神经网络，使用MNIST数据集来预测手写数字，但现在我遇到了一个难题，就是如何将自己的图像输入到CNN中进行预测。我已经保存了CNN训练后的权重，并希望使用它来预测我的图像。（注意：我的输入图像大小为28x28）

代码:

new_mnist.py :

ap = argparse.ArgumentParser()
ap.add_argument("-s", "--save-model", type=int, default=-1,
help="(optional) whether or not model should be saved to disk")  
ap.add_argument("-l", "--load-model", type=int, default=-1,
help="(optional) whether or not pre-trained model should be loaded")
ap.add_argument("-w", "--weights", type=str,
help="(optional) path to weights file")
args  = vars(ap.parse_args())

# fix random seed for reproducibility
seed = 7
numpy.random.seed(seed)

# load data
print("[INFO] downloading data...")
(X_train, y_train), (X_test, y_test) = mnist.load_data()
# reshape to be [samples][pixels][width][height]
X_train = X_train.reshape(X_train.shape[0], 1, 28, 28).astype('float32')
X_test = X_test.reshape(X_test.shape[0], 1, 28, 28).astype('float32')
print(X_test.shape[0])

# normalize inputs from 0-255 to 0-1
X_train = X_train / 255
X_test = X_test / 255
# one hot encode outputs
y_train = np_utils.to_categorical(y_train)
y_test = np_utils.to_categorical(y_test)
num_classes = y_test.shape[1]

# build the model
print("[INFO] compiling model...")
model = LeNet.build(num_classes = num_classes,weightsPath = args["weights"]          if args["load_model"] > 0 else None)
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

if args["load_model"] < 0:
# Fit the model
print("[INFO] training...")
model.fit(X_train, y_train, validation_data=(X_test, y_test), epochs=1,    batch_size=200, verbose=2)
# Final evaluation of the model
print("[INFO] evaluating...")
scores = model.evaluate(X_test, y_test, verbose=0)
print("Baseline Error: %.2f%%" % (100-scores[1]*100))
elif args["load_model"] > 0:
im = imread("C:\\Users\\Divyesh\\Desktop\\mnist.png")
im = im/255
pr = model.predict_classes(im)
print(pr)

# check to see if the model should be saved to file
if args["save_model"] > 0:
print("[INFO] dumping weights to file...")
model.save_weights(args["weights"], overwrite=True)

lenet.py :

class LeNet:
@staticmethod
def build(num_classes,weightsPath = None):
# create model
    model = Sequential()
    model.add(Convolution2D(30, 5, 5, border_mode='valid', input_shape=(1, 28, 28), activation='relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Convolution2D(15, 3, 3, activation='relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Dropout(0.2))
    model.add(Flatten())
    model.add(Dense(128, activation='relu'))
    model.add(Dense(50, activation='relu'))
    model.add(Dense(num_classes, activation='softmax'))
    # Compile model
    #model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
    if weightsPath is not None:
        model.load_weights(weightsPath)
    return model

在new_mnist.py中，我调用了predict(im)函数，其中im是一张28x28的图片，但是运行程序后，我收到了以下错误：

ValueError: Error when checking : expected conv2d_1_input to have 4      dimensions, but got array with shape (28, 28)

求助！！！

- thisisbhavin

Answer 1

尝试：

pr = model.predict_classes(im.reshape((1, 1, 28, 28)))

这里：第一个维度来自于样例（即使你只有一个样例也需要指定），第二个维度来自于通道（因为你使用 Theano 后端），剩下的是空间维度。

- Marcin Możejko

Answer 2

需要注意的是，图像必须以灰度方式上传。

1. 例如：

im = im[:,:,0]

2. 或者

import cv2

im = cv2.imread('C:\\Users\\Divyesh\\Desktop\\mnist.png', cv2.IMREAD_GRAYSCALE)

- Jean Marco