语义分割 - SegNet / DeconvNet
这里写目录标题
- 论文精读
-
- 先验知识
-
- 编码器 - 解码器结构（Encoder - Decoder frame）
- 随机丢弃层（Dropout）
- 反池化（unpooling）
- SegNet
-
- 本文创新点
- 算法架构
- DeconvNet
-
- 本文创新点
- 算法架构
- 算法实现
-
- SegNet.py
-
- 导入包
- 解码器中的卷积部分的定义
- 构建网络
- DeconvNet.py
-
- 导入包
- 解码器中反卷积的部分的定义
- 构建网络
论文精读
先验知识
编码器 - 解码器结构（Encoder - Decoder frame）
编码器结构：
由普通卷积层和下采样层将特征图尺寸缩小
解码器结构：
由普通卷积、上采样层和融合层组成，用来逐渐恢复空间维度，在尽可能减少信息损失的前提下完成同尺寸输入输出
随机丢弃层（Dropout）
当一个复杂的前馈神经网络被训练在小数据集上时，容易造成过拟合。Dropout通过忽略特征检测器来减少过拟合现象。
反池化（unpooling）
解码器中每一个最大池化层的索引都存储起来，用于之后在解码器中进行反池化操作。
- 有助于保持高频信息的完整性
- 对低分辨率的特征图进行反池化时，会忽略临近的信息
SegNet
本文创新点
- SegNet 核心的训练引擎包括一个 encoder 网络和一个对称的 decoder 网络，即编码器 - 解码器结构
- decoder 中直接利用与之对应的 encoder 阶段中最大池化时保留的 pooling index 来反池化
算法架构
DeconvNet
本文创新点
- 编码部分使用 VGG-16 卷积层
- 解码部分使用反卷积和反池化
- 特殊方法：将 object proposal 送入训练后的网络，整幅图像是由 proposal 分割结果的组合，解决了物体太大或太小带来的分割问题
算法架构
DeconvNet 与 SegNet 的不同之处主要在于
- 反池化与反卷积结合来形成 decoder
- encoder 和 decoder 之间加入了全连接
算法实现
除了模型以外，其他部分与FCN无异
SegNet.py
导入包
import torch
import torchvision.models as models
from torch import nn
解码器中的卷积部分的定义
def decoder(input_channel, output_channel, num=3):
if num == 3:
decoder_body = nn.Sequential(
nn.Conv2d(input_channel, input_channel, 3, padding=1),
nn.Conv2d(input_channel, input_channel, 3, padding=1),
nn.Conv2d(input_channel, output_channel, 3, padding=1))
elif num == 2:
decoder_body = nn.Sequential(
nn.Conv2d(input_channel, input_channel, 3, padding=1),
nn.Conv2d(input_channel, output_channel, 3, padding=1))
return decoder_body
构建网络
torch.nn.MaxPool2d() 用法
torch.nn.MaxUnpool2d() 用法
vgg16_pretrained = models.vgg16(pretrained=True)
class VGG16_seg(torch.nn.Module):
def __init__(self):
super(VGG16_seg, self).__init__()
pool_list = [4, 9, 16, 23, 30]
for index in pool_list:
vgg16_pretrained.features[index].return_indices = True # 让pooling层返回索引
self.encoder1 = vgg16_pretrained.features[:4]
self.pool1 = vgg16_pretrained.features[4]
self.encoder2 = vgg16_pretrained.features[5:9]
self.pool2 = vgg16_pretrained.features[9]
self.encoder3 = vgg16_pretrained.features[10:16]
self.pool3 = vgg16_pretrained.features[16]
self.encoder4 = vgg16_pretrained.features[17:23]
self.pool4 = vgg16_pretrained.features[23]
self.encoder5 = vgg16_pretrained.features[24:30]
self.pool5 = vgg16_pretrained.features[30]
self.decoder5 = decoder(512, 512)
self.unpool5 = nn.MaxUnpool2d(2, 2)
self.decoder4 = decoder(512, 256)
self.unpool4 = nn.MaxUnpool2d(2, 2)
self.decoder3 = decoder(256, 128)
self.unpool3 = nn.MaxUnpool2d(2, 2)
self.decoder2 = decoder(128, 64, 2)
self.unpool2 = nn.MaxUnpool2d(2, 2)
self.decoder1 = decoder(64, 12, 2)
self.unpool1 = nn.MaxUnpool2d(2, 2)
def forward(self, x): # 3, 352, 480
encoder1 = self.encoder1(x) # 64, 352, 480
output_size1 = encoder1.size() # 64, 352, 480
pool1, indices1 = self.pool1(encoder1) # 64, 176, 240
encoder2 = self.encoder2(pool1) # 128, 176, 240
output_size2 = encoder2.size() # 128, 176, 240
pool2, indices2 = self.pool2(encoder2) # 128, 88, 120
encoder3 = self.encoder3(pool2) # 256, 88, 120
output_size3 = encoder3.size() # 256, 88, 120
pool3, indices3 = self.pool3(encoder3) # 256, 44, 60
encoder4 = self.encoder4(pool3) # 512, 44, 60
output_size4 = encoder4.size() # 512, 44, 60
pool4, indices4 = self.pool4(encoder4) # 512, 22, 30
encoder5 = self.encoder5(pool4) # 512, 22, 30
output_size5 = encoder5.size() # 512, 22, 30
pool5, indices5 = self.pool5(encoder5) # 512, 11, 15
unpool5 = self.unpool5(input=pool5, indices=indices5, output_size=output_size5) # 512, 22, 30
decoder5 = self.decoder5(unpool5) # 512, 22, 30
unpool4 = self.unpool4(input=decoder5, indices=indices4, output_size=output_size4) # 512, 44, 60
decoder4 = self.decoder4(unpool4) # 256, 44, 60
unpool3 = self.unpool3(input=decoder4, indices=indices3, output_size=output_size3) # 256, 88, 120
decoder3 = self.decoder3(unpool3) # 128, 88, 120
unpool2 = self.unpool2(input=decoder3, indices=indices2, output_size=output_size2) # 128, 176, 240
decoder2 = self.decoder2(unpool2) # 64, 176, 240
unpool1 = self.unpool1(input=decoder2, indices=indices1, output_size=output_size1) # 64, 352, 480
decoder1 = self.decoder1(unpool1) # 12, 352, 480
return decoder1
DeconvNet.py
除了结构中加入了
导入包
import torch
import torchvision.models as models
from torch import nn
解码器中反卷积的部分的定义
与 SegNet 的不同点之一就是 decoder 部分采用了反卷积
def decoder(input_channel, output_channel, num=3):
if num == 3:
decoder_body = nn.Sequential(
nn.ConvTranspose2d(input_channel, input_channel, 3, padding=1),
nn.ConvTranspose2d(input_channel, input_channel, 3, padding=1),
nn.ConvTranspose2d(input_channel, output_channel, 3, padding=1))
elif num == 2:
decoder_body = nn.Sequential(
nn.ConvTranspose2d(input_channel, input_channel, 3, padding=1),
nn.ConvTranspose2d(input_channel, output_channel, 3, padding=1))
return decoder_body
构建网络
构建网络中，与 SegNet 不同点在于其加入了全连接部分，并在全连接后将数据 reshape 成了解码器的结束时的图片尺寸
class VGG16_deconv(torch.nn.Module):
def __init__(self):
super(VGG16_deconv, self).__init__()
pool_list = [4, 9, 16, 23, 30]
for index in pool_list:
vgg16_pretrained.features[index].return_indices = True
self.encoder1 = vgg16_pretrained.features[:4]
self.pool1 = vgg16_pretrained.features[4]
self.encoder2 = vgg16_pretrained.features[5:9]
self.pool2 = vgg16_pretrained.features[9]
self.encoder3 = vgg16_pretrained.features[10:16]
self.pool3 = vgg16_pretrained.features[16]
self.encoder4 = vgg16_pretrained.features[17:23]
self.pool4 = vgg16_pretrained.features[23]
self.encoder5 = vgg16_pretrained.features[24:30]
self.pool5 = vgg16_pretrained.features[30]
self.classifier = nn.Sequential(
torch.nn.Linear(512 * 11 * 15, 4096),
torch.nn.ReLU(),
torch.nn.Linear(4096, 512 * 11 * 15),
torch.nn.ReLU(),
)
self.decoder5 = decoder(512, 512)
self.unpool5 = nn.MaxUnpool2d(2, 2)
self.decoder4 = decoder(512, 256)
self.unpool4 = nn.MaxUnpool2d(2, 2)
self.decoder3 = decoder(256, 128)
self.unpool3 = nn.MaxUnpool2d(2, 2)
self.decoder2 = decoder(128, 64, 2)
self.unpool2 = nn.MaxUnpool2d(2, 2)
self.decoder1 = decoder(64, 12, 2)
self.unpool1 = nn.MaxUnpool2d(2, 2)
def forward(self, x): # 3, 352, 480
encoder1 = self.encoder1(x) # 64, 352, 480
output_size1 = encoder1.size() # 64, 352, 480
pool1, indices1 = self.pool1(encoder1) # 64, 176, 240
encoder2 = self.encoder2(pool1) # 128, 176, 240
output_size2 = encoder2.size() # 128, 176, 240
pool2, indices2 = self.pool2(encoder2) # 128, 88, 120
encoder3 = self.encoder3(pool2) # 256, 88, 120
output_size3 = encoder3.size() # 256, 88, 120
pool3, indices3 = self.pool3(encoder3) # 256, 44, 60
encoder4 = self.encoder4(pool3) # 512, 44, 60
output_size4 = encoder4.size() # 512, 44, 60
pool4, indices4 = self.pool4(encoder4) # 512, 22, 30
encoder5 = self.encoder5(pool4) # 512, 22, 30
output_size5 = encoder5.size() # 512, 22, 30
pool5, indices5 = self.pool5(encoder5) # 512, 11, 15
# 在输入全连接前要将其打平成(batchsize, -1)
pool5 = pool5.view(pool5.size(0), -1)
fc = self.classifier(pool5) # 在全连接结束后返回的也是(batchsize, -1)的打平形式
fc = fc.reshape(1, 512, 11, 15)
unpool5 = self.unpool5(input=fc, indices=indices5, output_size=output_size5) # 512, 22, 30
decoder5 = self.decoder5(unpool5) # 512, 22, 30
unpool4 = self.unpool4(input=decoder5, indices=indices4, output_size=output_size4) # 512, 44, 60
decoder4 = self.decoder4(unpool4) # 256, 44, 60
unpool3 = self.unpool3(input=decoder4, indices=indices3, output_size=output_size3) # 256, 88, 120
decoder3 = self.decoder3(unpool3) # 128, 88, 120
unpool2 = self.unpool2(input=decoder3, indices=indices2, output_size=output_size2) # 128, 176, 240
decoder2 = self.decoder2(unpool2) # 64, 176, 240
unpool1 = self.unpool1(input=decoder2, indices=indices1, output_size=output_size1) # 64, 352, 480
decoder1 = self.decoder1(unpool1) # 12, 352, 480
return decoder1