keras-lenet5

from keras.layers import Input, Conv2D, MaxPooling2D, Dense, Flatten
from keras.models import Model
from keras.optimizers import Adam
from keras.utils import np_utils
from keras.callbacks import ModelCheckpoint
import numpy as np
import pandas as pd

img_row = 28
img_col = 28

#读取训练集数据，图片数据格式是(batch, row*col)，但是我们在训练的过程中输入是(batch,row,col,channel),所以我们需要处理成指定的格式。
def create_train(path):
    img_matrix = pd.read_csv(path).values
    total = img_matrix.shape[0]
    labels = img_matrix[:, 0]
    x_train = img_matrix[:, 1:]
    return labels.astype(np.uint8), x_train.reshape(total, img_row, img_col, 1)


def create_test(path):
    img_matrix = pd.read_csv(path).values
    total = img_matrix.shape[0]
    return img_matrix.reshape(total, img_row, img_col, 1)


def create_lenet():
    inputs = Input((img_row, img_col, 1))
    conv1 = Conv2D(filters=6, kernel_size=(5, 5), strides=1, padding='valid', activation='relu')(inputs)
    pool1 = MaxPooling2D(pool_size=(2, 2), strides=None, padding='valid')(conv1)

    conv2 = Conv2D(filters=16, kernel_size=(5, 5), strides=1, padding='valid', activation='relu')(pool1)
    pool2 = MaxPooling2D(pool_size=(2, 2), strides=None, padding='valid')(conv2)

    conv3 = Conv2D(filters=120, kernel_size=(1, 1), strides=1, padding='valid')(pool2)

    full1 = Flatten()(conv3)#卷积层是(batch,rows,cols,channel),为了全连接需要将其展开得到(batch,rows*cols*channel)的平铺层
    full2 = Dense(units=84)(full1)
    outputs = Dense(units=10, activation='softmax')(full2)

    model = Model(inputs=[inputs], outputs=[outputs])

    model.compile(optimizer=Adam(), loss='categorical_crossentropy', metrics=['accuracy'])

    return model


def train_and_predict():
    print('-' * 30)
    print('Loading and preprocessing train data...')
    print('-' * 30)

    labels, x_train = create_train('dataset/Digital/train.csv')
    # Make the value floats in [0;1] instead of int in [0;255]
    x_train = x_train.astype('float32')
    x_train /= 255

    # convert class vectors to binary class matrices (ie one-hot vectors)
    labels = np_utils.to_categorical(labels, 10)
    print('-' * 30)
    print('Creating and compiling model...')
    print('-' * 30)

    lenet5 = create_lenet()
    model_checkpoint = ModelCheckpoint('weights.h5', monitor='val_loss', save_best_only=True)

    print('-' * 30)
    print('Fitting model...')
    print('-' * 30)

    lenet5.fit(x_train, labels, batch_size=32, epochs=20, verbose=1, shuffle=True, validation_split=0.2,
               callbacks=[model_checkpoint])

    print('-' * 30)
    print('Loading and preprocessing test data...')
    print('-' * 30)
    x_test = create_test('dataset/Digital/test.csv')
    x_test = x_test.astype('float32')
    x_test /= 255
    print('-' * 30)
    print('Loading saved weights...')
    print('-' * 30)
    lenet5.load_weights('weights.h5')

    print('-' * 30)
    print('Predicting test data...')
    print('-' * 30)
    y_pred = lenet5.predict(x_test, verbose=1)
    np.savetxt('mnist-pred.csv', np.c_[range(1, len(y_pred) + 1), y_pred], delimiter=',', header='ImageId,Label',
               comments='',
               fmt='%d')

    print('-' * 30)
    print('Saving predicted masks to files...')
    print('-' * 30)


if __name__ == '__main__':
    train_and_predict()