def test():
check_weight()
model = cnn_models()
model.load_weights(WEIGHT_PATH)
test_x, test_y = get_pictures(False)
out = model.predict_on_batch(test_x)
print("loss = {}, accuracy = {}".format(out[0], out[1]))
def test():
print("begin testing")
check_weight(CNN_WEIGHT_PATH)
model = cnn_models()
model.load_weights(CNN_WEIGHT_PATH)
test_x, test_y = get_pictures(Is_Train=False)
out = model.test_on_batch(test_x, test_y)
print("loss = {:.4f}, accuracy = {:.4f}".format(out[0], out[1]))
print("test finish\n")
def train(Is_continue=True):
batch_x, batch_y = get_pictures()
model = cnn_models()
if Is_continue:
if os.path.exists(CNN_WEIGHT_PATH):
model.load_weights(CNN_WEIGHT_PATH)
else:
print("begin training!")
# callback回调函数会在训练适当时机被调用 validation_split=0.1在训练集中随机选取10%做验证集（可以理解为测试集）
model.fit(x=batch_x, y=batch_y, batch_size=batch_size, epochs=epochs, verbose=2, validation_split=0.1)
model.save_weights(CNN_WEIGHT_PATH, True) # 保存模型的权重
print("train finish\n")
def train():
batch_x, batch_y = get_pictures()
model = cnn_models()
if os.path.exists(WEIGHT_PATH):
model.load_weights(WEIGHT_PATH)
#tb = TensorBoard(log_dir=LOG_PATH, write_images=1)
#call_list = [tb]
print("begin training!")
model.fit(x=batch_x,
y=batch_y,
batch_size=batch_size,
epochs=epochs,
verbose=2) # callback回调函数会在训练适当时机被调用
model.save_weights(WEIGHT_PATH, True)