# 无人驾驶汽车系统入门：深度前馈网络，深度学习的正则化，交通信号识别

▌深度学习的能力

ImageNet是一个拥有1400万张图片的巨大数据集，基于ImageNet数据集，ILSVRC（ImageNet Large Scale Visual Recognition Challenge）挑战赛每年举办一次。

▌深度前馈神经网络——为什么要深？

▌应用于深度神经网络的正则化技术

Dropout

defexpend_training_data(train_x, train_y):

"""

Augment training data

"""

expanded_images = np.zeros([train_x.shape[] *5, train_x.shape[1], train_x.shape[2]])

expanded_labels = np.zeros([train_x.shape[] *5])

counter =

forx, yinzip(train_x, train_y):

# register original data

expanded_images[counter, :, :] = x

expanded_labels[counter] = y

counter = counter +1

# get a value for the background

# zero is the expected value, but median() is used to estimate background's value

bg_value = np.median(x)# this is regarded as background's value

foriinrange(4):

# rotate the image with random degree

angle = np.random.randint(-15,15,1)

new_img = ndimage.rotate(x, angle, reshape=False, cval=bg_value)

# shift the image with random distance

shift = np.random.randint(-2,2,2)

new_img_ = ndimage.shift(new_img, shift, cval=bg_value)

# register new training data

expanded_images[counter, :, :] = new_img_

expanded_labels[counter] = y

counter = counter +1

returnexpanded_images, expanded_labels

agument_x, agument_y = expend_training_data(x_train[:3], y_train[:3])

L2惩罚一方面降低了权重的学习自由度，削弱了网络的学习能力，另一方面相对均匀的权重又能使模型光滑化，使模型对输入的细微变化不敏感，从而增强模型的鲁棒性。

Dropout

▌基于深度前馈神经网络的交通信号识别

Belgium Traffic Sign Dataset 数据集

http://btsd.ethz.ch/shareddata/BelgiumTSC/BelgiumTSC_Training.zip

• 发表于:
• 原文链接https://kuaibao.qq.com/s/20180602B0R4PZ00?refer=cp_1026
• 腾讯「云+社区」是腾讯内容开放平台帐号（企鹅号）传播渠道之一，根据《腾讯内容开放平台服务协议》转载发布内容。
• 如有侵权，请联系 yunjia_community@tencent.com 删除。

2021-12-05

2021-12-05

2021-12-05

2021-12-05

2021-12-05

2021-12-05

2018-07-05

2021-12-05

2021-12-05

2021-12-05

2021-12-05

2021-12-05