tf = py.importlib.import_module('tensorflow');
np = py.importlib.import_module('numpy');
plt = py.importlib.import_module('matplotlib.pyplot');
sio = py.importlib.import_module('scipy.io');
run = py.importlib.import_module('run');
py.importlib.reload(run);
% 载入数据
loaddata = run.load_data();
% train_images = double(loaddata{1}); % 载入训练数据
% train_labels = double(loaddata{2});
test_images = double(loaddata{3}); % 载入测试数据
test_labels  = double(loaddata{4});
% 载入训练好的模型
model = run.create_model();
model.load_weights("training_1/cp.ckpt");
%% 使用
idx_sel = [1:39];
image_use = cell([1,length(idx_sel)]);
labels_use = zeros(1,length(idx_sel));
for i = 1:20% 测试用例：如果从matlab传参到直接调用keras模型中的fit方法会报错，image_use = np.array(reshape(test_images(idx_sel(i),:,:),[1,28,28]));label_true = test_labels(idx_sel(i));% 执行预测label_pred = int64(model.predict_classes(image_use));% 显示结果subplot(4,5,i)image_use_matlab = reshape(double(image_use),[28,28]); % 转换成matlab格式的imageimshow(image_use_matlab)if label_pred == label_truetitle(['预测',num2str(label_pred),'，真实',num2str(label_true)],'color','g')    elsetitle(['预测',num2str(label_pred),'，真实',num2str(label_true)],'color','r')    end
end

import tensorflow as tf
from tensorflow import keras
import numpy as np
import scipy.io as sio
#import matplotlib.pyplot as plt
import osdef create_model():model = keras.Sequential([keras.layers.Flatten(input_shape=(28, 28)),keras.layers.Dense(128, activation=tf.nn.relu),keras.layers.Dense(10, activation=tf.nn.softmax)])model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['accuracy'])    return modeldef load_data():path='./mnist.npz'      f = np.load(path)  with np.load(path) as f:train_images, train_labels = f['x_train'], f['y_train']  test_images, test_labels = f['x_test'], f['y_test']      return train_images, train_labels, test_images, test_labelsdef train():train_images, train_labels, test_images, test_labels = load_data()    train_images = np.array(train_images)train_labels = np.array(train_labels)test_images = np.array(test_images)test_labels = np.array(test_labels)train_images = train_images / 255.0test_images = test_images / 255.0#plt.figure(figsize=(10,10))#for i in range(25):#    plt.subplot(5,5,i+1)#    plt.xticks([])#    plt.yticks([])#    plt.grid(False)#    plt.imshow(train_images[i], cmap=plt.cm.binary)#    plt.xlabel(class_names[train_labels[i]])#plt.show()checkpoint_path = "training_1/cp.ckpt"checkpoint_dir = os.path.dirname(checkpoint_path)# Create checkpoint callbackcp_callback = tf.keras.callbacks.ModelCheckpoint(checkpoint_path,save_weights_only=True,verbose=1)model = create_model()model.fit(train_images, train_labels, epochs=5,callbacks=[cp_callback])test_loss, test_acc = model.evaluate(test_images, test_labels)print('Test accuracy:', test_acc)predictions = model.predict(test_images)return modelif __name__ == '__main__':train_images, train_labels, test_images, test_labels = load_data()model = train()