一，手提包数据集

数据集下载：用pytorch写FCN进行手提包的语义分割。

training data(https://github.com/yunlongdong/FCN-pytorch-easiest/tree/master/last)，放到bag_data文件夹下
ground-truth label(https://github.com/yunlongdong/FCN-pytorch-easiest/tree/master/last_msk)，放到bag_data_mask文件夹下

项目目录结构：

训练数据：

训练label：

从这个手提包数据集可以看出，这是个二分类的，就是只分割出手提包和背景两个类别。所以label处黑色的表示手提包，白色的就是无关的背景。

二，训练代码（用来读取数据集，包括手提包图片和手提包图片的label）

2.1：数据集读取的代码

###BagData.py
import os
import torch
import torch.nn as nn
from torch.utils.data import DataLoader, Dataset, random_split
from torchvision import transforms
import numpy as np
import cv2
#from onehot import onehot

transform = transforms.Compose([
    transforms.ToTensor(), 
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])

def onehot(data, n):
    buf = np.zeros(data.shape + (n, ))
    nmsk = np.arange(data.size)*n + data.ravel()
    buf.ravel()[nmsk-1] = 1
    return buf

class BagDataset(Dataset):
    def __init__(self, transform=None):
        self.transform = transform
        
    def __len__(self):
        return len(os.listdir('bag_data'))

    def __getitem__(self, idx):
        img_name = os.listdir('bag_data')[idx]
        imgA = cv2.imread('bag_data/'+img_name)
        imgA = cv2.resize(imgA, (160, 160))
        #print(imgA.shape)
        imgB = cv2.imread('bag_data_msk/'+img_name, 0)
        imgB = cv2.resize(imgB, (160, 160))
        #print(imgB.shape)
        imgB = imgB/255
        imgB = imgB.astype('uint8')
        imgB = onehot(imgB, 2)    #因为此代码是二分类问题，即分割出手提包和背景两样就行，因此这里参数是2
        imgB = imgB.transpose(2,0,1) #imgB不经过transform处理，所以要手动把(H,W,C)转成(C,H,W)
        imgB = torch.FloatTensor(imgB)
        if self.transform:
            imgA = self.transform(imgA) #一转成向量后，imgA通道就变成(C,H,W)
        return imgA, imgB

bag = BagDataset(transform)

train_size = int(0.9 * len(bag))    #整个训练集中，百分之90为训练集
test_size = len(bag) - train_size
train_dataset, test_dataset = random_split(bag, [train_size, test_size]) #划分训练集和测试集

train_dataloader = DataLoader(train_dataset, batch_size=4, shuffle=True, num_workers=4)
test_dataloader = DataLoader(test_dataset, batch_size=4, shuffle=True, num_workers=4)


if __name__ =='__main__':
    for train_batch in train_dataloader:
        print(train_batch)

    for test_batch in test_dataloader:
        print(test_batch)

贴了数据集读取的代码后，我觉得有必要说一下onehot这个函数。

1.就是数据集label的onehot化：

onehot化是把label化成一个一维向量。

onehot化的函数如下：

def onehot(data, n):
    buf = np.zeros(data.shape + (n, ))
    nmsk = np.arange(data.size)*n + data.ravel()
    buf.ravel()[nmsk-1] = 1
    return buf

输入的data为以灰度图形式读取的label，n为分割的类别数(此数据集是2)

buf = np.zeros(data.shape + (n, ))#设data的shape为(a,b)，则构造一个全0矩阵，维度为(a,b,n)

因为n是2，所以意思就是，2层的(a，b)的全0矩阵，一层用来表示手提包的，剩下一层则是用来表示背景的。

nmsk = np.arange(data.size)*n + data.ravel()

这行则比较妙一点，首先设data的size为5，则arange(5)为，(0，1，2，3，4)，其实就是表示data各个元素的位置。arange(5)*2为(0，2，4，6，8)，其实这是变相表示原来长度x2的位置。而data因为是label，且归一化过的，所以data里的值要么是0要么是1，data.ravel()是把data展成一维数组，arange(5)*2+data.ravel()意思是在(0，2，4，6，8)中，表示手提包的则+1，表示背景的则+0。这里打个比方，例如第三个和第五个位置是表示手提包的，则是(0，2，5，6，9)，到这里可能还看不出什么，结合下一句代码就明白了。

buf.ravel()[nmsk-1] = 1

用回刚刚的例子(0，2，5，6，9)，nmsk-1后，是(9，1，4，5，8)，与初始的(0，2，4，6，8)对比，若原来是1的位置会保持原样(因为+1后又-1了)，而原本是0的，表示其位置就会-1。这样的结果就是把(a,b)的label投射到(a,b)*2的长度中。这样做的原因数据集是2分类的，所以网络输出肯定是(a,b,2)这样的，所以label必须要和网络输出维度形式一样才能比较，得出损失函数。

2.2，模型代码

#####FCN.py
import torch
import torch.nn as nn
from torchvision import models
from torchvision.models.vgg import VGG


class FCNs(nn.Module):

    def __init__(self, pretrained_net, n_class):
        super().__init__()
        self.n_class = n_class
        self.pretrained_net = pretrained_net
        self.relu    = nn.ReLU(inplace=True)
        self.deconv1 = nn.ConvTranspose2d(512, 512, kernel_size=3, stride=2, padding=1, dilation=1, output_padding=1)
        self.bn1     = nn.BatchNorm2d(512)
        self.deconv2 = nn.ConvTranspose2d(512, 256, kernel_size=3, stride=2, padding=1, dilation=1, output_padding=1)
        self.bn2     = nn.BatchNorm2d(256)
        self.deconv3 = nn.ConvTranspose2d(256, 128, kernel_size=3, stride=2, padding=1, dilation=1, output_padding=1)
        self.bn3     = nn.BatchNorm2d(128)
        self.deconv4 = nn.ConvTranspose2d(128, 64, kernel_size=3, stride=2, padding=1, dilation=1, output_padding=1)
        self.bn4     = nn.BatchNorm2d(64)
        self.deconv5 = nn.ConvTranspose2d(64, 32, kernel_size=3, stride=2, padding=1, dilation=1, output_padding=1)
        self.bn5     = nn.BatchNorm2d(32)
        self.classifier = nn.Conv2d(32, n_class, kernel_size=1) 
        # classifier is 1x1 conv, to reduce channels from 32 to n_class

    def forward(self, x):
        output = self.pretrained_net(x)
        x5 = output['x5']  
        x4 = output['x4']  
        x3 = output['x3']  
        x2 = output['x2']  
        x1 = output['x1']  

        score = self.bn1(self.relu(self.deconv1(x5)))     
        score = score + x4                                
        score = self.bn2(self.relu(self.deconv2(score)))  
        score = score + x3                                
        score = self.bn3(self.relu(self.deconv3(score)))  
        score = score + x2                                
        score = self.bn4(self.relu(self.deconv4(score)))  
        score = score + x1                                
        score = self.bn5(self.relu(self.deconv5(score)))  
        score = self.classifier(score)                    

        return score  


class VGGNet(VGG):
    def __init__(self, pretrained=False, model='vgg16', requires_grad=True, remove_fc=True, show_params=False):
        super().__init__(make_layers(cfg[model]))
        self.ranges = ranges[model]

        if pretrained:
            exec("self.load_state_dict(models.%s(pretrained=False).state_dict())" % model)

        if not requires_grad:
            for param in super().parameters():
                param.requires_grad = False

        # delete redundant fully-connected layer params, can save memory
        # 去掉vgg最后的全连接层(classifier)
        if remove_fc:  
            del self.classifier

        if show_params:
            for name, param in self.named_parameters():
                print(name, param.size())

    def forward(self, x):
        output = {}
        # get the output of each maxpooling layer (5 maxpool in VGG net)
        for idx, (begin, end) in enumerate(self.ranges):
        #self.ranges = ((0, 5), (5, 10), (10, 17), (17, 24), (24, 31)) (vgg16 examples)
            for layer in range(begin, end):
                x = self.features[layer](x)
            output["x%d"%(idx+1)] = x

        return output


ranges = {
    'vgg11': ((0, 3), (3, 6),  (6, 11),  (11, 16), (16, 21)),
    'vgg13': ((0, 5), (5, 10), (10, 15), (15, 20), (20, 25)),
    'vgg16': ((0, 5), (5, 10), (10, 17), (17, 24), (24, 31)),
    'vgg19': ((0, 5), (5, 10), (10, 19), (19, 28), (28, 37))
}

# Vgg-Net config 
# Vgg网络结构配置
cfg = {
    'vgg11': [64, 'M', 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
    'vgg13': [64, 64, 'M', 128, 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
    'vgg16': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512, 'M', 512, 512, 512, 'M'],
    'vgg19': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 256, 'M', 512, 512, 512, 512, 'M', 512, 512, 512, 512, 'M'],
}

# make layers using Vgg-Net config(cfg)
# 由cfg构建vgg-Net
def make_layers(cfg, batch_norm=False):
    layers = []
    in_channels = 3
    for v in cfg:
        if v == 'M':
            layers += [nn.MaxPool2d(kernel_size=2, stride=2)]
        else:
            conv2d = nn.Conv2d(in_channels, v, kernel_size=3, padding=1)
            if batch_norm:
                layers += [conv2d, nn.BatchNorm2d(v), nn.ReLU(inplace=True)]
            else:
                layers += [conv2d, nn.ReLU(inplace=True)]
            in_channels = v
    return nn.Sequential(*layers)

'''
VGG-16网络参数

Sequential(
  (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (1): ReLU(inplace)
  (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (3): ReLU(inplace)
  (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (6): ReLU(inplace)
  (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (8): ReLU(inplace)
  (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (11): ReLU(inplace)
  (12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (13): ReLU(inplace)
  (14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (15): ReLU(inplace)
  (16): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (18): ReLU(inplace)
  (19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (20): ReLU(inplace)
  (21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (22): ReLU(inplace)
  (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (25): ReLU(inplace)
  (26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (27): ReLU(inplace)
  (28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (29): ReLU(inplace)
  (30): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)

'''


if __name__ == "__main__":
    pass

2.3。train代码

########train.py
from datetime import datetime

import matplotlib.pyplot as plt
import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
import visdom

from BagData import test_dataloader, train_dataloader
from FCN import FCN8s, FCN16s, FCN32s, FCNs, VGGNet


def train(epo_num=50, show_vgg_params=False):

    #vis = visdom.Visdom()

    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

    vgg_model = VGGNet(requires_grad=True, show_params=show_vgg_params)
    fcn_model = FCNs(pretrained_net=vgg_model, n_class=2)
    fcn_model = fcn_model.to(device)
    criterion = nn.BCELoss().to(device)
    optimizer = optim.SGD(fcn_model.parameters(), lr=1e-2, momentum=0.7)

    all_train_iter_loss = []
    all_test_iter_loss = []

    # start timing
    prev_time = datetime.now()
    for epo in range(epo_num):
        
        train_loss = 0
        fcn_model.train()
        for index, (bag, bag_msk) in enumerate(train_dataloader):
            # bag.shape is torch.Size([4, 3, 160, 160])
            # bag_msk.shape is torch.Size([4, 2, 160, 160])

            bag = bag.to(device)
            bag_msk = bag_msk.to(device)

            optimizer.zero_grad()
            output = fcn_model(bag)
            output = torch.sigmoid(output) # output.shape is torch.Size([4, 2, 160, 160])
            # print(output)
            # print(bag_msk)
            loss = criterion(output, bag_msk)
            loss.backward()
            iter_loss = loss.item()
            all_train_iter_loss.append(iter_loss)
            train_loss += iter_loss
            optimizer.step()

            output_np = output.cpu().detach().numpy().copy() # output_np.shape = (4, 2, 160, 160)  
            output_np = np.argmin(output_np, axis=1)
            bag_msk_np = bag_msk.cpu().detach().numpy().copy() # bag_msk_np.shape = (4, 2, 160, 160) 
            bag_msk_np = np.argmin(bag_msk_np, axis=1)

        test_loss = 0
        fcn_model.eval()
        with torch.no_grad():
            for index, (bag, bag_msk) in enumerate(test_dataloader):

                bag = bag.to(device)
                bag_msk = bag_msk.to(device)

                optimizer.zero_grad()
                output = fcn_model(bag)
                output = torch.sigmoid(output) # output.shape is torch.Size([4, 2, 160, 160])
                loss = criterion(output, bag_msk)
                iter_loss = loss.item()
                all_test_iter_loss.append(iter_loss)
                test_loss += iter_loss

                output_np = output.cpu().detach().numpy().copy() # output_np.shape = (4, 2, 160, 160)  
                output_np = np.argmin(output_np, axis=1)
                bag_msk_np = bag_msk.cpu().detach().numpy().copy() # bag_msk_np.shape = (4, 2, 160, 160) 
                bag_msk_np = np.argmin(bag_msk_np, axis=1)
        
        cur_time = datetime.now()
        h, remainder = divmod((cur_time - prev_time).seconds, 3600)
        m, s = divmod(remainder, 60)
        time_str = "Time %02d:%02d:%02d" % (h, m, s)
        prev_time = cur_time

        print('epoch train loss = %f, epoch test loss = %f, %s'
                %(train_loss/len(train_dataloader), test_loss/len(test_dataloader), time_str))
        

        if np.mod(epo, 5) == 0:
            torch.save(fcn_model, 'checkpoints/fcn_model_{}.pt'.format(epo))
            print('saveing checkpoints/fcn_model_{}.pt'.format(epo))


if __name__ == "__main__":
    train(epo_num=100, show_vgg_params=False)

三，预测代码：

#######t.py

import matplotlib.pyplot as plt
import numpy as np
import torch
from torchvision import transforms
import os
import cv2
import matplotlib.pyplot as plt
from BagData import test_dataloader, train_dataloader
from FCN import FCN8s, FCN16s, FCN32s, FCNs, VGGNet

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = torch.load('checkpoints/fcn_model_95.pt')  # 加载模型
model = model.to(device)
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])


if __name__ =='__main__':
    img_name =r'bag3.jpg'  #预测的图片
    imgA = cv2.imread(img_name)
    imgA = cv2.resize(imgA, (160, 160))

    imgA = transform(imgA)
    imgA = imgA.to(device)
    imgA = imgA.unsqueeze(0)
    output = model(imgA)
    output = torch.sigmoid(output)

    output_np = output.cpu().detach().numpy().copy()  # output_np.shape = (4, 2, 160, 160)
    print(output_np.shape)   #(1, 2, 160, 160)
    output_np = np.argmin(output_np, axis=1)
    print(output_np.shape)  #(1,160, 160)

    plt.subplot(1, 2, 1)
    #plt.imshow(np.squeeze(bag_msk_np[0, ...]), 'gray')
    #plt.subplot(1, 2, 2)
    plt.imshow(np.squeeze(output_np[0, ...]), 'gray')
    plt.pause(3)