VGG+SegNet模型

最新推荐文章于 2022-07-11 16:51:12 发布
最新推荐文章于 2022-07-11 16:51:12 发布
阅读量374 收藏 2
点赞数
分类专栏： # 机器学习之Keras 文章标签： 深度学习 图像语义分割 VggSegNet 卷积神经网络 TensorFlow Keras
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/LYJ20010728/article/details/115498149
版权

搭建模型结构代码

from tensorflow.keras import layers
from tensorflow.keras import models


def VggSegNet(input_shape, num_classes):
    # 输入图片的尺寸
    inputs = layers.Input(shape=input_shape)
    
    # 512,512,3 -> 512,512,64
    conv_1 = layers.Conv2D(filters=64, kernel_size=(3,3), strides=(1,1), padding='same')(inputs)
    conv_1 = layers.BatchNormalization()(conv_1)
    conv_1 = layers.Activation('relu')(conv_1)
    conv_2 = layers.Conv2D(filters=64, kernel_size=(3,3), strides=(1,1), padding='same')(conv_1)
    conv_2 = layers.BatchNormalization()(conv_2)
    conv_2 = layers.Activation('relu')(conv_2)
    
    # 512,512,64 -> 256,256,64
    pool_1 = layers.MaxPooling2D(pool_size=(2,2))(conv_2)

    # 256,256,64 -> 256,256,128
    conv_3 = layers.Conv2D(filters=128, kernel_size=(3,3), strides=(1,1), padding='same')(pool_1)
    conv_3 = layers.BatchNormalization()(conv_3)
    conv_3 = layers.Activation('relu')(conv_3)
    conv_4 = layers.Conv2D(filters=128, kernel_size=(3,3), strides=(1,1), padding='same')(conv_3)
    conv_4 = layers.BatchNormalization()(conv_4)
    conv_4 = layers.Activation('relu')(conv_4)

    # 256,256,128 -> 128,128,128
    pool_2 = layers.MaxPooling2D(pool_size=(2,2))(conv_4)

    # 128,128,128 -> 128,128,256
    conv_5 = layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), padding='same')(pool_2)
    conv_5 = layers.BatchNormalization()(conv_5)
    conv_5 = layers.Activation('relu')(conv_5)
    conv_6 = layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), padding='same')(conv_5)
    conv_6 = layers.BatchNormalization()(conv_6)
    conv_6 = layers.Activation('relu')(conv_6)
    conv_7 = layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), padding='same')(conv_6)
    conv_7 = layers.BatchNormalization()(conv_7)
    conv_7 = layers.Activation('relu')(conv_7)

    # 128,128,256 -> 64,64,256
    pool_3 = layers.MaxPooling2D(pool_size=(2,2))(conv_7)

    # 64,64,256 -> 64,64,512
    conv_8 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding='same')(pool_3)
    conv_8 = layers.BatchNormalization()(conv_8)
    conv_8 = layers.Activation('relu')(conv_8)
    conv_9 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding='same')(conv_8)
    conv_9 = layers.BatchNormalization()(conv_9)
    conv_9 = layers.Activation('relu')(conv_9)
    conv_10 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1) ,padding='same')(conv_9)
    conv_10 = layers.BatchNormalization()(conv_10)
    conv_10 = layers.Activation('relu')(conv_10)

    # 64,64,512 -> 32,32,512
    pool_4 = layers.MaxPooling2D(pool_size=(2,2))(conv_10)

    # 32,32,512 -> 32,32,512
    conv_11 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding='same')(pool_4)
    conv_11 = layers.BatchNormalization()(conv_11)
    conv_11 = layers.Activation('relu')(conv_11)
    conv_12 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding='same')(conv_11)
    conv_12 = layers.BatchNormalization()(conv_12)
    conv_12 = layers.Activation('relu')(conv_12)
    conv_13 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding='same')(conv_12)
    conv_13 = layers.BatchNormalization()(conv_13)
    conv_13 = layers.Activation('relu')(conv_13)

    pool_5 = layers.MaxPooling2D(pool_size=(2,2))(conv_13)


    # 32,32,512 -> 64,64,512
    unpool_1 = layers.UpSampling2D(size=(2,2))(pool_5)

    # 64,64,512 -> 
    conv_14 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding="same")(unpool_1)
    conv_14 = layers.BatchNormalization()(conv_14)
    conv_14 = layers.Activation("relu")(conv_14)
    conv_15 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding="same")(conv_14)
    conv_15 = layers.BatchNormalization()(conv_15)
    conv_15 = layers.Activation("relu")(conv_15)
    conv_16 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding="same")(conv_15)
    conv_16 = layers.BatchNormalization()(conv_16)
    conv_16 = layers.Activation("relu")(conv_16)

    unpool_2 = layers.UpSampling2D(size=(2,2))(conv_16)

    conv_17 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding="same")(unpool_2)
    conv_17 = layers.BatchNormalization()(conv_17)
    conv_17 = layers.Activation("relu")(conv_17)
    conv_18 = layers.Conv2D(filters=512, kernel_size=(3,3), strides=(1,1), padding="same")(conv_17)
    conv_18 = layers.BatchNormalization()(conv_18)
    conv_18 = layers.Activation("relu")(conv_18)
    conv_19 = layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), padding="same")(conv_18)
    conv_19 = layers.BatchNormalization()(conv_19)
    conv_19 = layers.Activation("relu")(conv_19)

    unpool_3 = layers.MaxPooling2D(pool_size=(2,2))(conv_19)

    conv_20 = layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), padding="same")(unpool_3)
    conv_20 = layers.BatchNormalization()(conv_20)
    conv_20 = layers.Activation("relu")(conv_20)
    conv_21 = layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), padding="same")(conv_20)
    conv_21 = layers.BatchNormalization()(conv_21)
    conv_21 = layers.Activation("relu")(conv_21)
    conv_22 = layers.Conv2D(filters=128, kernel_size=(3,3), strides=(1,1), padding="same")(conv_21)
    conv_22 = layers.BatchNormalization()(conv_22)
    conv_22 = layers.Activation("relu")(conv_22)

    unpool_4 = layers.MaxPooling2D(pool_size=(2,2))(conv_22)

    conv_23 = layers.Conv2D(filters=128, kernel_size=(3,3), strides=(1,1), padding="same")(unpool_4)
    conv_23 = layers.BatchNormalization()(conv_23)
    conv_23 = layers.Activation("relu")(conv_23)
    conv_24 = layers.Conv2D(64, kernel_size=(3,3), strides=(1,1), padding="same")(conv_23)
    conv_24 = layers.BatchNormalization()(conv_24)
    conv_24 = layers.Activation("relu")(conv_24)

    unpool_5 = layers.MaxPooling2D(pool_size=(2,2))(conv_24)

    conv_25 = layers.Conv2D(filters=64, kernel_size=(3,3), strides=(1,1), padding="same")(unpool_5)
    conv_25 = layers.BatchNormalization()(conv_25)
    conv_25 = layers.Activation("relu")(conv_25)

    conv_26 = layers.Conv2D(filters=num_classes, kernel_size=(1, 1), strides=(1,1), padding="valid")(conv_25)
    conv_26 = layers.BatchNormalization()(conv_26)
#     conv_26 = Reshape(
#         (input_shape[0] * input_shape[1], n_labels),
#         input_shape=(input_shape[0], input_shape[1], n_labels),
#     )(conv_26)

    outputs = layers.Activation('softmax')(conv_26)
    
    model = models.Model(inputs=inputs, outputs=outputs)
    return model

测试模型并的出模型结构层次

input_size = (512,512,3)
num_classes = 21
model = VggSegNet(input_size, num_classes)
model.summary()

Model: "model"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 512, 512, 3)]     0         
_________________________________________________________________
conv2d (Conv2D)              (None, 512, 512, 64)      1792      
_________________________________________________________________
batch_normalization (BatchNo (None, 512, 512, 64)      256       
_________________________________________________________________
activation (Activation)      (None, 512, 512, 64)      0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 512, 512, 64)      36928     
_________________________________________________________________
batch_normalization_1 (Batch (None, 512, 512, 64)      256       
_________________________________________________________________
activation_1 (Activation)    (None, 512, 512, 64)      0         
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 256, 256, 64)      0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 256, 256, 128)     73856     
_________________________________________________________________
batch_normalization_2 (Batch (None, 256, 256, 128)     512       
_________________________________________________________________
activation_2 (Activation)    (None, 256, 256, 128)     0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 256, 256, 128)     147584    
_________________________________________________________________
batch_normalization_3 (Batch (None, 256, 256, 128)     512       
_________________________________________________________________
activation_3 (Activation)    (None, 256, 256, 128)     0         
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 128, 128, 128)     0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 128, 128, 256)     295168    
_________________________________________________________________
batch_normalization_4 (Batch (None, 128, 128, 256)     1024      
_________________________________________________________________
activation_4 (Activation)    (None, 128, 128, 256)     0         
_________________________________________________________________
conv2d_5 (Conv2D)            (None, 128, 128, 256)     590080    
_________________________________________________________________
batch_normalization_5 (Batch (None, 128, 128, 256)     1024      
_________________________________________________________________
activation_5 (Activation)    (None, 128, 128, 256)     0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 128, 128, 256)     590080    
_________________________________________________________________
batch_normalization_6 (Batch (None, 128, 128, 256)     1024      
_________________________________________________________________
activation_6 (Activation)    (None, 128, 128, 256)     0         
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 64, 64, 256)       0         
_________________________________________________________________
conv2d_7 (Conv2D)            (None, 64, 64, 512)       1180160   
_________________________________________________________________
batch_normalization_7 (Batch (None, 64, 64, 512)       2048      
_________________________________________________________________
activation_7 (Activation)    (None, 64, 64, 512)       0         
_________________________________________________________________
conv2d_8 (Conv2D)            (None, 64, 64, 512)       2359808   
_________________________________________________________________
batch_normalization_8 (Batch (None, 64, 64, 512)       2048      
_________________________________________________________________
activation_8 (Activation)    (None, 64, 64, 512)       0         
_________________________________________________________________
conv2d_9 (Conv2D)            (None, 64, 64, 512)       2359808   
_________________________________________________________________
batch_normalization_9 (Batch (None, 64, 64, 512)       2048      
_________________________________________________________________
activation_9 (Activation)    (None, 64, 64, 512)       0         
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 32, 32, 512)       0         
_________________________________________________________________
conv2d_10 (Conv2D)           (None, 32, 32, 512)       2359808   
_________________________________________________________________
batch_normalization_10 (Batc (None, 32, 32, 512)       2048      
_________________________________________________________________
activation_10 (Activation)   (None, 32, 32, 512)       0         
_________________________________________________________________
conv2d_11 (Conv2D)           (None, 32, 32, 512)       2359808   
_________________________________________________________________
batch_normalization_11 (Batc (None, 32, 32, 512)       2048      
_________________________________________________________________
activation_11 (Activation)   (None, 32, 32, 512)       0         
_________________________________________________________________
conv2d_12 (Conv2D)           (None, 32, 32, 512)       2359808   
_________________________________________________________________
batch_normalization_12 (Batc (None, 32, 32, 512)       2048      
_________________________________________________________________
activation_12 (Activation)   (None, 32, 32, 512)       0         
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 16, 16, 512)       0         
_________________________________________________________________
up_sampling2d (UpSampling2D) (None, 32, 32, 512)       0         
_________________________________________________________________
conv2d_13 (Conv2D)           (None, 32, 32, 512)       2359808   
_________________________________________________________________
batch_normalization_13 (Batc (None, 32, 32, 512)       2048      
_________________________________________________________________
activation_13 (Activation)   (None, 32, 32, 512)       0         
_________________________________________________________________
conv2d_14 (Conv2D)           (None, 32, 32, 512)       2359808   
_________________________________________________________________
batch_normalization_14 (Batc (None, 32, 32, 512)       2048      
_________________________________________________________________
activation_14 (Activation)   (None, 32, 32, 512)       0         
_________________________________________________________________
conv2d_15 (Conv2D)           (None, 32, 32, 512)       2359808   
_________________________________________________________________
batch_normalization_15 (Batc (None, 32, 32, 512)       2048      
_________________________________________________________________
activation_15 (Activation)   (None, 32, 32, 512)       0         
_________________________________________________________________
up_sampling2d_1 (UpSampling2 (None, 64, 64, 512)       0         
_________________________________________________________________
conv2d_16 (Conv2D)           (None, 64, 64, 512)       2359808   
_________________________________________________________________
batch_normalization_16 (Batc (None, 64, 64, 512)       2048      
_________________________________________________________________
activation_16 (Activation)   (None, 64, 64, 512)       0         
_________________________________________________________________
conv2d_17 (Conv2D)           (None, 64, 64, 512)       2359808   
_________________________________________________________________
batch_normalization_17 (Batc (None, 64, 64, 512)       2048      
_________________________________________________________________
activation_17 (Activation)   (None, 64, 64, 512)       0         
_________________________________________________________________
conv2d_18 (Conv2D)           (None, 64, 64, 256)       1179904   
_________________________________________________________________
batch_normalization_18 (Batc (None, 64, 64, 256)       1024      
_________________________________________________________________
activation_18 (Activation)   (None, 64, 64, 256)       0         
_________________________________________________________________
max_pooling2d_5 (MaxPooling2 (None, 32, 32, 256)       0         
_________________________________________________________________
conv2d_19 (Conv2D)           (None, 32, 32, 256)       590080    
_________________________________________________________________
batch_normalization_19 (Batc (None, 32, 32, 256)       1024      
_________________________________________________________________
activation_19 (Activation)   (None, 32, 32, 256)       0         
_________________________________________________________________
conv2d_20 (Conv2D)           (None, 32, 32, 256)       590080    
_________________________________________________________________
batch_normalization_20 (Batc (None, 32, 32, 256)       1024      
_________________________________________________________________
activation_20 (Activation)   (None, 32, 32, 256)       0         
_________________________________________________________________
conv2d_21 (Conv2D)           (None, 32, 32, 128)       295040    
_________________________________________________________________
batch_normalization_21 (Batc (None, 32, 32, 128)       512       
_________________________________________________________________
activation_21 (Activation)   (None, 32, 32, 128)       0         
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 16, 16, 128)       0         
_________________________________________________________________
conv2d_22 (Conv2D)           (None, 16, 16, 128)       147584    
_________________________________________________________________
batch_normalization_22 (Batc (None, 16, 16, 128)       512       
_________________________________________________________________
activation_22 (Activation)   (None, 16, 16, 128)       0         
_________________________________________________________________
conv2d_23 (Conv2D)           (None, 16, 16, 64)        73792     
_________________________________________________________________
batch_normalization_23 (Batc (None, 16, 16, 64)        256       
_________________________________________________________________
activation_23 (Activation)   (None, 16, 16, 64)        0         
_________________________________________________________________
max_pooling2d_7 (MaxPooling2 (None, 8, 8, 64)          0         
_________________________________________________________________
conv2d_24 (Conv2D)           (None, 8, 8, 64)          36928     
_________________________________________________________________
batch_normalization_24 (Batc (None, 8, 8, 64)          256       
_________________________________________________________________
activation_24 (Activation)   (None, 8, 8, 64)          0         
_________________________________________________________________
conv2d_25 (Conv2D)           (None, 8, 8, 21)          1365      
_________________________________________________________________
batch_normalization_25 (Batc (None, 8, 8, 21)          84        
_________________________________________________________________
activation_25 (Activation)   (None, 8, 8, 21)          0         
=================================================================
Total params: 29,460,329
Trainable params: 29,444,415
Non-trainable params: 15,914
_________________________________________________________________
H3rmesk1t
关注
专栏目录
FCN相关问题的一些整理 (FCN VGG Segnet resnet )
有些代码不应该被忘记
08-18 1万+
1  FCN 相对CNN的优点 1) 2014年，加州大学伯克利分校的Long等人提出的完全卷积网络(Fully Convolutional Networks)，推广了原有的CNN结构，在不带有全连接层的情况下能进行密集预测。 这种结构的提出使得分割图谱可以生成任意大小的图像，且与图像块分类方法相比，也提高了处理速度。在后来，几乎所有关于语义分割的最新研究都采用了这种结构。 2) 除
VGGNET理解
dulingtingzi的博客
04-06 7589
    此文章大部分引用了   大数据与人工智能Lab“微信号（BigdataAILab） 里面对VGGNET部分的文章，感觉这个公众号很不错，里面的白话深度学习部分非常受用。在读的过程中遇到了很多不懂的问题，通过各种查资料搞明白了很多，记录下来，给和我一样的小伙伴们提供一个比较快的途径，也方便自己以后复习。         本文的结构主要围绕这篇VGGNET的文章展开，里面加了自己的理解的部分...
基于VGG16主干模型的segnet语义分割详解及实例
mcu旺仔acm的博客
07-11 1852
基于VGG16主干模型的segnet语义分割详解及实例
【Keras】基于SegNet和U-Net的遥感图像语义分割-附件资源
03-05
【Keras】基于SegNet和U-Net的遥感图像语义分割-附件资源
SE_VGG
Major_S的博客
08-24 1549
import torch.nn as nn import torch class SE_VGG(nn.Module): def __init__(self, num_classes): super().__init__() self.num_classes = num_classes # define an empty for Conv_ReLU_MaxPool net = [] # block 1
CVPR2018_主干网络系列(3)_SE-Net及其衍生的网络[SE-XX-Net]
kevin_zhao_zl的博客
04-22 3715
该论文聚焦通道维度的优化提出一个名为SE(Squeeze-and-Excitation)模块的新颖的单元架构，利用通道的独立性分别重调通道维度的特征来获得更好的最终效果，这些模块堆叠而成的新网络最终也取得了非常好的效果。
基于VGGnet模型的segnet斑马线划分
03-07
1、利用VGGnet提取图片特征 2、利用segnet进行decoder 3、使用tensorflow里面kereas里面神经网络八步法，一步一步教你训练自己的模型 4、提供了用于训练和验证的数据集 5、网络注释清晰，方便二次开发和修改
SegNet_ResNet_resnet语义分割_segnet_语义分割resnet_迁移学习
09-11
总的来说，这个项目展示了如何利用深度学习和迁移学习技术，特别是ResNet和SegNet模型，来解决语义分割问题。这种技术对于自动驾驶、医疗影像分析、遥感图像处理等领域具有巨大的应用潜力。理解并掌握这些技术对于...
SegNet+tensorflow代码+数据集.rar
08-23
在这个项目中，你将找到一个使用TensorFlow实现的SegNet模型，以及相关的训练数据集。 编码器部分是基于VGG16的预训练网络，这是一个在ImageNet数据集上训练的深度模型，具有16个层。VGG16的编码阶段通过卷积和池化...
SegNet+dataset
06-04
`Tensorflow-SegNet.rar`可能是使用TensorFlow框架实现的SegNet模型代码或预训练权重。`SegNet.zip`可能包含了SegNet模型的源代码或者模型结构的详细描述。"新建文件夹"和"新建文件夹(1)"可能包含的是数据集的子集，...
图像语义分割网络：SegNet
09-28
发表SegNet网络的论文为：Badrinarayanan V, Kendall A, Cipolla R.SegNet: A Deep Convolutional Encoder-Decoder Architecture for SceneSegmentation[J]. IEEE Transactions on Pattern Analysis & MachineIntelligence, 2017, PP(99):1-1。 来源于美国加州大学伯克利分校的这项工作为语义分割引入了端到端的全卷积网络，在构建的网络结构中，重新利用ImageNet的预训练网络用于语义分割，并使用了反卷积层进行上采样并且引入跳跃连接来改善上采样粗糙的像素定位。
keras-segmentation:开始使用基于Keras的语义分割，包括FCN32FCN8SegNetU-Net
05-11
实施不是原始文件。 该项目的目的是开始语义分割并掌握基本过程。 发布FCN32 / 8，SegNet，U-Net谢谢您的支持。 环境 物品 价值 物品 价值 凯拉斯 2.2.4 作业系统 赢10 张量流gpu 1.10 / 1.12 Python 3.6.7 参考 数据： 数据或： 项目计划书 python visualizeDataset.py ：可视示例 python train.py ：执行火车python predict.py预测python predict.py ：执行预测 您可以在项目切换模型中修改参数或克隆历史版本。 关于 FCN32 可视化结果： FCN8 可视化结果： 隔离网 网络
Resnet50.h5
12-17
直接下载通过调用里面参数及模型结构就可以快速训练好模型，帮助你节省时间！
Segnet网络结构
10-26
Segnet网络结构
u-net 实现(keras)
04-17
利用keras实现u-net这样一个全卷积神经网络，进行图像分割
SegNet网络结构
流浪若相惜的专栏
06-04 1万+
@SegNet网络结构 刚刚接触深度学习–semantic segmentation相关的研究，对SegNet的网络结构进行了学习，虽然已经有了很多的解释，还是想要自己写一下，将整体结构做一个梳理。博客底部附有参考链接，感谢大神们的分析以及代码的赞助。 SegNet是通过对图像中每一个像素点进行分类，识别每一个像素点的类别来实现图像的分割。其思路与FCN的思路十分相似，只是两者的Encoder与D...
【语义分割】--SegNet理解
热门推荐
Alex-zzm
03-27 6万+
原文地址：SegNet复现详解：http://mi.eng.cam.ac.uk/projects/segnet/tutorial.html 实现代码:    github                       TensorFlow简介：        SegNet是Cambridge提出旨在解决自动驾驶或者智能机器人的图像语义分割深度网络，开放源码，基于caffe框架。SegNet基于FCN...
keras实现Segnet、FCN、U-Net系列分割网络
qq_29462849的博客
12-02 6585
介绍 语义分割是计算机视觉领域中非常重要的一个领域，在无人驾驶、图像分析、图像理解领域中有很多应用。在这里就不再详细叙述语义分割算法，具体算法请参考以下链接： Segnet：https://blog.csdn.net/zhuzemin45/article/details/79709874 U-Net： https://blog.csdn.net/m_buddy/article/details/79...
SegNet：深度卷积编码-解码器架构用于图像分割
SegNet网络结构是深度学习领域中用于语义分割的重要模型。语义分割是一项计算机视觉任务，其目标是将图像中的每个像素分配到预定义的类别中，如行人、车辆、建筑物等。SegNet的设计理念是通过深度卷积神经网络实现...
评论 1
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值