如何将测试仪的准确度从85％提高到90％？

Question

我使用Keras CNN模型遇到二进制图像分类问题（总共约700张图像），现在其训练准确度约为95％，测试准确度约为85％，如何提高测试准确度？

我已经尝试过Hyperas超参数优化，并使用OpenCV生成经过预处理的阈值图像，使其最多可容纳约4.5K数据，以及Keras API中的数据增强，但这是我能得到的最好的结果。

model = Sequential()
model.add(Conv2D(32, (3, 3), input_shape=input_shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(32, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(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'))

optimizer = keras.optimizers.RMSprop(lr = 0.0002,
                    rho = 0.95,
                    epsilon = 1e-07,
                    decay = 0.00001
)

model.compile(loss = 'binary_crossentropy',
              optimizer = optimizer,
              #optimizer = 'rmsprop',
              metrics = ['accuracy'])

test_datagen = ImageDataGenerator(rescale=1. / 255)

train_datagen = ImageDataGenerator(
    ### Newly added 05.14.
    featurewise_center = True, 
    samplewise_center = True, 
    featurewise_std_normalization = True, 
    samplewise_std_normalization = True, 
    zca_whitening = True, 
    zca_epsilon = 1e-06,
    channel_shift_range = 0.1,
    ###

    rotation_range = 360, ### Used to be 180
    width_shift_range = 0.2,
    height_shift_range = 0.2,
    brightness_range = (0.8, 1.2),
    rescale = 1. / 255,
    shear_range = 0.2,
    zoom_range = 0.2,
    horizontal_flip = True,
    vertical_flip = True)

train_generator = train_datagen.flow_from_directory(
    train_data_dir,
    target_size = (img_width, img_height),
    batch_size = batch_size,
    class_mode ='binary')

validation_generator = test_datagen.flow_from_directory(
    validation_data_dir,
    target_size = (img_width, img_height),
    batch_size = batch_size,
    class_mode = 'binary')

history = model.fit_generator(
    train_generator,
    steps_per_epoch = nb_train_samples // batch_size,
    epochs = epochs,
    validation_data = validation_generator,
    validation_steps = nb_validation_samples // batch_size
    #callbacks = callbacks
)

基本上，我希望测试数据的准确性达到90％以上，并保持现在的训练准确性。