def MCU_output():
'''
输出混淆矩阵
:return:
'''
dir = MCU_dir
model = My_CNN(40, 17, 1000, dir) # 初始化模型
model.LBCNN('Adam')
path = dir + 'LBCNN_model.h5'
model.run_model(path)
def load_img():
'''
载入图片，只需要做一次
:return:
'''
# MCU
dir = MCU_dir
model = My_CNN(64, 17, 1000, dir, channel=1) # 初始化模型
#img_dir = '' # 图片保存路径
model.read_img(img_dir) # 初次使用需要载入图片
def Lenet_MCU():
'''
:return:
'''
dir = MCU_dir # 数据集保存路径，之后就可以不用重新载入图片
model = My_CNN(64, 17, 1000, dir, channel=1) # 初始化模型
model.LeNet_5('Adam', lr=1e-3)
hist_path = dir + 'LeNet_'
path = dir + 'LeNet_model.h5'
model.train_model(1000, 512, model_path=path, hist_path=hist_path, early_save=True, patience=10)
def LBCNN_MFPT():
'''
使用LBCNN完整地训练数据集，保存最优结果
:return:
'''
dir = MFPT_dir
model = My_CNN(40, 16, 1000, dir) # 初始化模型
model.LBCNN('Adam')
hist_path = dir + 'LBCNN_'
path = dir + 'LBCNN_model.h5'
model.train_model(1000,
512,
model_path=path,
hist_path=hist_path,
early_save=True)
def Lenet_MFPT():
'''
使用Lenet完整地训练数据集，保存最优结果
:return:
'''
dir = MFPT_dir
model = My_CNN(40, 16, 1000, dir) # 初始化模型
model.LeNet_5('Adam', lr=1e-3)
hist_path = dir + 'LeNet_'
path = dir + 'LeNet_model.h5'
model.train_model(1000,
512,
model_path=path,
hist_path=hist_path,
early_save=True)
def Lenet_MCU():
'''
使用Lenet完整地训练数据集，保存最优结果
:return:
'''
dir = MCU_dir # 数据集保存路径
model = My_CNN(40, 17, 1000, dir) # 初始化模型
model.LeNet_5('Adam', lr=1e-4)
hist_path = dir + 'LeNet_'
path = dir + 'LeNet_model.h5'
model.train_model(1000,
512,
model_path=path,
hist_path=hist_path,
early_save=True,
patience=100)
def LBCNN_MCU():
'''
使用LBCNN完整地训练数据集，保存最优结果
:return:
'''
dir = MCU_dir
model = My_CNN(40, 17, 1000, dir) # 初始化模型
model.LBCNN('Adam', lr=1e-4)
hist_path = dir + 'LBCNN_'
path = dir + 'LBCNN_model.h5'
model.train_model(1000,
512,
model_path=path,
hist_path=hist_path,
early_save=True,
patience=100)
def load_img():
'''
载入图片，只需要做一次
:return:
'''
# MFPT
dir = MFPT_dir
model = My_CNN(40, 16, 1000, dir) # 初始化模型
img_dir = 'imgs\\' # 图片保存路径
model.read_img(img_dir)
# MCU
dir = MCU_dir
model = My_CNN(40, 17, 1000, dir) # 初始化模型
img_dir = 'F:\毕业设计\MCU\时频图\imgs\\' # 图片保存路径
model.read_img(img_dir)
def MCU_race():
'''
比较LBCNN和Lenet在数据集中的表现
:return:
'''
dir = MCU_dir # 数据集保存路径
model = My_CNN(40, 17, 1000, dir) # 初始化模型
model.LeNet_5('Adam', lr=1e-4)
hist_path = dir + 'RLeNet_'
path = dir + 'RLeNet_model.h5'
model.train_model(30,
512,
model_path=path,
hist_path=hist_path,
early_save=False)
model.LBCNN('Adam', lr=1e-4)
hist_path = dir + 'RLBCNN_'
path = dir + 'RLBCNN_model.h5'
model.train_model(30,
512,
model_path=path,
hist_path=hist_path,
early_save=False)
def plot():
'''
画图
:return:
'''
dir = MCU_dir # 数据集保存路径
model = My_CNN(40, 17, 1000, dir) # 初始化模型
hist_path = dir + 'Adam_'
history = model.load_data(hist_path, 'history')
plt.figure()
plt.plot(history['acc'])
hist_path = dir + 'SGD_'
history = model.load_data(hist_path, 'history')
plt.plot(history['acc'])
hist_path = dir + 'Adadelta_'
history = model.load_data(hist_path, 'history')
plt.plot(history['acc'])
hist_path = dir + 'Adagrad_'
history = model.load_data(hist_path, 'history')
plt.plot(history['acc'])
plt.legend(['Adam', 'SGD', 'Adadelta', 'Adagrad'])
plt.show()
def OP_MCU():
'''
对数据集，比较不同优化器下的LBCNN网络的效果
:return: 模型会自动保存为H5格式，训练历史数据会自动保存为二进制格式
'''
dir = MCU_dir
model = My_CNN(40, 17, 1000, dir) # 初始化模型
# Adam
model.LBCNN('Adam')
hist_path = dir + 'Adam_'
path = dir + 'Adam_model.h5'
model.train_model(30, 512, model_path=path, hist_path=hist_path)
# SGD
model.LBCNN('SGD')
hist_path = dir + 'SGD_'
path = dir + 'SGD_model.h5'
model.train_model(30, 512, model_path=path, hist_path=hist_path)
# Adagrad
model.LBCNN('Adagrad')
hist_path = dir + 'Adagrad_'
path = dir + 'Adagrad_model.h5'
model.train_model(30, 512, model_path=path, hist_path=hist_path)
# Adadelta
model.LBCNN('Adadelta')
hist_path = dir + 'Adadelta_'
path = dir + 'Adadelta_model.h5'
model.train_model(30, 512, model_path=path, hist_path=hist_path)
def write():
# 将二进制文件写为csv格式，方便之后用MATLAB画图
# MFPT
dir = MFPT_dir # 数据集保存路径
model = My_CNN(40, 16, 1000, dir) # 初始化模型
# LBCNN+Adam
hist_path = dir + 'Adam_'
history = model.load_data(hist_path, 'history')
filename = dir + 'Adam.csv'
csv_trans(history, filename)
# LBCNN+SGD
hist_path = dir + 'SGD_'
history = model.load_data(hist_path, 'history')
filename = dir + 'SGD.csv'
csv_trans(history, filename)
# LBCNN+Adagrad
hist_path = dir + 'Adagrad_'
history = model.load_data(hist_path, 'history')
filename = dir + 'Adagrad.csv'
csv_trans(history, filename)
# LBCNN+Adadelta
hist_path = dir + 'Adadelta_'
history = model.load_data(hist_path, 'history')
filename = dir + 'Adadelta.csv'
csv_trans(history, filename)
#LBCNN
hist_path = dir + 'LBCNN_'
history = model.load_data(hist_path, 'history')
filename = dir + 'LBCNN.csv'
csv_trans(history, filename)
# Lenet
hist_path = dir + 'Lenet_'
history = model.load_data(hist_path, 'history')
filename = dir + 'Lenet.csv'
csv_trans(history, filename)
# race
hist_path = dir + 'RLenet_'
history = model.load_data(hist_path, 'history')
filename = dir + 'RLenet.csv'
csv_trans(history, filename)
hist_path = dir + 'RLBCNN_'
history = model.load_data(hist_path, 'history')
filename = dir + 'RLBCNN.csv'
csv_trans(history, filename)
# MCU
dir = MCU_dir # 数据集保存路径
model = My_CNN(40, 17, 1000, dir) # 初始化模型
# LBCNN+Adam
hist_path = dir + 'Adam_'
history = model.load_data(hist_path, 'history')
filename = dir + 'Adam.csv'
csv_trans(history, filename)
# LBCNN+SGD
hist_path = dir + 'SGD_'
history = model.load_data(hist_path, 'history')
filename = dir + 'SGD.csv'
csv_trans(history, filename)
# LBCNN+Adagrad
hist_path = dir + 'Adagrad_'
history = model.load_data(hist_path, 'history')
filename = dir + 'Adagrad.csv'
csv_trans(history, filename)
# LBCNN+Adadelta
hist_path = dir + 'Adadelta_'
history = model.load_data(hist_path, 'history')
filename = dir + 'Adadelta.csv'
csv_trans(history, filename)
# LBCNN
hist_path = dir + 'LBCNN_'
history = model.load_data(hist_path, 'history')
filename = dir + 'LBCNN.csv'
csv_trans(history, filename)
# Lenet
hist_path = dir + 'Lenet_'
history = model.load_data(hist_path, 'history')
filename = dir + 'Lenet.csv'
csv_trans(history, filename)
# race
hist_path = dir + 'RLenet_'
history = model.load_data(hist_path, 'history')
filename = dir + 'RLenet.csv'
csv_trans(history, filename)
hist_path = dir + 'RLBCNN_'
history = model.load_data(hist_path, 'history')
filename = dir + 'RLBCNN.csv'
csv_trans(history, filename)