语义分割系列5-Pspnet（pytorch实现）
Pspnet全名Pyramid Scene Parsing Network，论文地址：Pyramid Scene Parsing Network
论文名就是《Pyramid Scene Parsing Network》。
该模型提出是为了解决场景分析问题。针对FCN网络在场景分析数据集上存在的问题，Pspnet提出一系列改进方案，以提升场景分析中对于相似颜色、形状的物体的检测精度。
图1 ADE20k场景分析
作者在ADE20K数据集上进行实验时，主要发现有如下3个问题：
- 错误匹配，FCN模型把水里的船预测成汽车，但是汽车是不会在水上的。因此，作者认为FCN缺乏收集上下文能力，导致了分类错误。
- 作者发现相似的标签会导致一些奇怪的错误，比如earth和field，mountain和hill，wall，house，building和skyscraper。FCN模型会出现混淆。
- 第三是小目标的丢失问题，像一些路灯、信号牌这种小物体，很难被FCN所发现。相反的，一些特别大的物体预测中，在感受野不够大的情况下，往往会丢失一部分信息，导致预测不连续。
为了解决这些问题，作者提出了Pyramid Pooling Module。
Pyramid Pooling Module
作者在文章中提出了Pyramid Pooling Module（池化金字塔结构）这一模块。
作者提到，在深层网络中，感受野的大小大致上体现了模型能获得的上下文新消息。尽管在理论上Resnet的感受野已经大于图像尺寸，但是实际上会小得多。这就导致了很多网络不能充分的将上下文信息结合起来，于是作者就提出了一种全局的先验方法-全局平均池化。
作者在PPM模块中并联了四个不同大小的全局池化层，将原始的feature map池化生成不同级别的特征图，经过卷积和上采样恢复到原始大小。这种操作聚合了多尺度的图像特征，生成了一个“hierarchical global prior”，融合了不同尺度和不同子区域之间的信息。最后，这个先验信息再和原始特征图进行相加，输入到最后的卷积模块完成预测。
图2 Pspnet
Pspnet的核心就是PPM模块。其网络架构十分简单，backbone为resnet网络，将原始图像下采样8倍成特征图，特征图输入到PPM模块，并与其输出相加，最后经过卷积和8倍双线性差值上采样得到结果（图2）。
辅助损失
图3 辅助损失
论文中还有一个细节是辅助损失（auxiliary loss），在resnet101的res4b22层引出一条FCN分支，用于计算辅助损失。论文里设置了赋值损失loss2的权重为0.4。则最终的损失则为：
论文复现
本文主要在CamVid数据集上进行复现，数据集可以在另一篇博客中找到CamVid数据集的创建和使用。
Resnet
这里调用了pytorch官方写的ResNet101，替换最后两个layer为dialation模式，只采用8倍下采样。引出layer3的计算结果用于计算辅助损失。
from torchvision.models import resnet50, resnet101
from torchvision.models._utils import IntermediateLayerGetter
import torch
import torch.nn as nn
backbone=IntermediateLayerGetter(
resnet101(pretrained=False, replace_stride_with_dilation=[False, True, True]),
return_layers={'layer3':'aux','layer4': 'stage4'}
)
x = torch.randn(1, 3, 224, 224).cpu()
result = backbone(x)
for k, v in result.items():
print(k, v.shape)pspnet
from torchvision.models import resnet50, resnet101
from torchvision.models._utils import IntermediateLayerGetter
import torch
import torch.nn as nn
class PPM(nn.ModuleList):
def __init__(self, pool_sizes, in_channels, out_channels):
super(PPM, self).__init__()
self.pool_sizes = pool_sizes
self.in_channels = in_channels
self.out_channels = out_channels
for pool_size in pool_sizes:
self.append(
nn.Sequential(
nn.AdaptiveMaxPool2d(pool_size),
nn.Conv2d(self.in_channels, self.out_channels, kernel_size=1),
)
)
def forward(self, x):
out_puts = []
for ppm in self:
ppm_out = nn.functional.interpolate(ppm(x), size=x.size()[-2:], mode='bilinear', align_corners=True)
out_puts.append(ppm_out)
return out_puts
class PSPHEAD(nn.Module):
def __init__(self, in_channels, out_channels,pool_sizes = [1, 2, 3, 6],num_classes=3):
super(PSPHEAD, self).__init__()
self.pool_sizes = pool_sizes
self.num_classes = num_classes
self.in_channels = in_channels
self.out_channels = out_channels
self.psp_modules = PPM(self.pool_sizes, self.in_channels, self.out_channels)
self.final = nn.Sequential(
nn.Conv2d(self.in_channels + len(self.pool_sizes)*self.out_channels, self.out_channels, kernel_size=3, padding=1),
nn.BatchNorm2d(self.out_channels),
nn.ReLU(),
)
def forward(self, x):
out = self.psp_modules(x)
out.append(x)
out = torch.cat(out, 1)
out = self.final(out)
return out
# 构建一个FCN分割头，用于计算辅助损失
class Aux_Head(nn.Module):
def __init__(self, in_channels=1024, num_classes=3):
super(Aux_Head, self).__init__()
self.num_classes = num_classes
self.in_channels = in_channels
self.decode_head = nn.Sequential(
nn.Conv2d(self.in_channels, self.in_channels//2, kernel_size=3, padding=1),
nn.BatchNorm2d(self.in_channels//2),
nn.ReLU(),
nn.Conv2d(self.in_channels//2, self.in_channels//4, kernel_size=3, padding=1),
nn.BatchNorm2d(self.in_channels//4),
nn.ReLU(),
nn.Conv2d(self.in_channels//4, self.num_classes, kernel_size=3, padding=1),
)
def forward(self, x):
return self.decode_head(x)
class Pspnet(nn.Module):
def __init__(self, num_classes, aux_loss = True):
super(Pspnet, self).__init__()
self.num_classes = num_classes
self.backbone = IntermediateLayerGetter(
resnet50(pretrained=False, replace_stride_with_dilation=[False, True, True]),
return_layers={'layer3':"aux" ,'layer4': 'stage4'}
)
self.aux_loss = aux_loss
self.decoder = PSPHEAD(in_channels=2048, out_channels=512, pool_sizes = [1, 2, 3, 6], num_classes=self.num_classes)
self.cls_seg = nn.Sequential(
nn.Conv2d(512, self.num_classes, kernel_size=3, padding=1),
)
if self.aux_loss:
self.aux_head = Aux_Head(in_channels=1024, num_classes=self.num_classes)
def forward(self, x):
_, _, h, w = x.size()
feats = self.backbone(x)
x = self.decoder(feats["stage4"])
x = self.cls_seg(x)
x = nn.functional.interpolate(x, size=(h, w),mode='bilinear', align_corners=True)
# 如果需要添加辅助损失
if self.aux_loss:
aux_output = self.aux_head(feats['aux'])
aux_output = nn.functional.interpolate(aux_output, size=(h, w),mode='bilinear', align_corners=True)
return {"output":x, "aux_output":aux_output}
return {"output":x}
if __name__ == "__main__":
model = Pspnet(num_classes=3, aux_loss=True)
model = model.cuda()
a = torch.ones([2, 3, 224, 224])
a = a.cuda()
for name, out in model(a).items():
print(name,": ", out.shape)数据集构建
# 导入库
import os
os.environ['CUDA_VISIBLE_DEVICES'] = '0'
import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
from torch import optim
from torch.utils.data import Dataset, DataLoader, random_split
from tqdm import tqdm
import warnings
warnings.filterwarnings("ignore")
import os.path as osp
import matplotlib.pyplot as plt
from PIL import Image
import numpy as np
import albumentations as A
from albumentations.pytorch.transforms import ToTensorV2
torch.manual_seed(17)
# 自定义数据集CamVidDataset
class CamVidDataset(torch.utils.data.Dataset):
"""CamVid Dataset. Read images, apply augmentation and preprocessing transformations.
Args:
images_dir (str): path to images folder
masks_dir (str): path to segmentation masks folder
class_values (list): values of classes to extract from segmentation mask
augmentation (albumentations.Compose): data transfromation pipeline
(e.g. flip, scale, etc.)
preprocessing (albumentations.Compose): data preprocessing
(e.g. noralization, shape manipulation, etc.)
"""
def __init__(self, images_dir, masks_dir):
self.transform = A.Compose([
A.Resize(448, 448),
A.HorizontalFlip(),
A.VerticalFlip(),
A.Normalize(),
ToTensorV2(),
])
self.ids = os.listdir(images_dir)
self.images_fps = [os.path.join(images_dir, image_id) for image_id in self.ids]
self.masks_fps = [os.path.join(masks_dir, image_id) for image_id in self.ids]
def __getitem__(self, i):
# read data
image = np.array(Image.open(self.images_fps[i]).convert('RGB'))
mask = np.array( Image.open(self.masks_fps[i]).convert('RGB'))
image = self.transform(image=image,mask=mask)
return image['image'], image['mask'][:,:,0]
def __len__(self):
return len(self.ids)
# 设置数据集路径
DATA_DIR = r'dataset\camvid' # 根据自己的路径来设置
x_train_dir = os.path.join(DATA_DIR, 'train_images')
y_train_dir = os.path.join(DATA_DIR, 'train_labels')
x_valid_dir = os.path.join(DATA_DIR, 'valid_images')
y_valid_dir = os.path.join(DATA_DIR, 'valid_labels')
train_dataset = CamVidDataset(
x_train_dir,
y_train_dir,
)
val_dataset = CamVidDataset(
x_valid_dir,
y_valid_dir,
)
train_loader = DataLoader(train_dataset, batch_size=8, shuffle=True,drop_last=True)
val_loader = DataLoader(val_dataset, batch_size=8, shuffle=True,drop_last=True)模型训练
from d2l import torch as d2l
from tqdm import tqdm
import pandas as pd
model = Pspnet(num_classes=32, aux_loss=True)
model = model.cuda()
# training loop 100 epochs
epochs_num = 100
# 选用SGD优化器来训练
optimizer = torch.optim.SGD(model.parameters(), lr=0.1)
schedule = torch.optim.lr_scheduler.StepLR(optimizer, step_size=50)
# 损失函数选用多分类交叉熵损失函数
lossf = nn.CrossEntropyLoss(ignore_index=255)
def evaluate(net, data_iter, device=torch.device('cuda:0')):
net.eval()
metric = d2l.Accumulator(3)
with torch.no_grad():
for X, y in data_iter:
if isinstance(X, list):
X = [x.to(device) for x in X]
else:
X = X.to(device)
y = y.to(device)
pred = net(X)['output']
metric.add(d2l.accuracy(pred, y), d2l.size(y))
return metric[0] / metric[1]
# 训练函数
def train_ch13(net, train_iter, test_iter, loss_func, optimizer, num_epochs, schedule, devices=d2l.try_all_gpus()):
timer, num_batches = d2l.Timer(), len(train_iter)
animator = d2l.Animator(xlabel='epoch', xlim=[1, num_epochs], ylim=[0, 1], legend=['train loss', 'train acc', 'test acc'])
net = nn.DataParallel(net, device_ids=devices).to(devices[0])
# 用来保存一些训练参数
loss_list = []
train_acc_list = []
test_acc_list = []
epochs_list = []
time_list = []
lr_list = []
for epoch in range(num_epochs):
# metric: loss, accuracy, labels.shape[0], labels.numel(), 0.4*aux_loss
metric = d2l.Accumulator(5)
for i, (X, labels) in enumerate(train_iter):
timer.start()
if isinstance(X, list):
X = [x.to(devices[0]) for x in X]
else:
X = X.to(devices[0])
gt = labels.long().to(devices[0])
net.train()
optimizer.zero_grad()
result = net(X)
seg_loss = loss_func(result['output'], gt)
aux_loss = loss_func(result['aux_output'], gt)
loss_sum = seg_loss + 0.4*aux_loss
l = loss_sum
loss_sum.sum().backward()
optimizer.step()
acc = d2l.accuracy(result['output'], gt)
metric.add(l, acc, labels.shape[0], labels.numel(), 0.4*aux_loss)
timer.stop()
if (i + 1) % (num_batches // 5) == 0 or i == num_batches - 1:
animator.add(epoch + (i + 1) / num_batches, (metric[0] / metric[2], metric[1] / metric[3], None, None))
test_acc = evaluate(net, test_iter)
animator.add(epoch + 1, (None, None, test_acc))
schedule.step()
print(f"epoch {epoch+1}/{epochs_num} --- loss {metric[0]/metric[2]:.3f} --- aux_loss {metric[4]/metric[2]:.3f} --- train acc {metric[1]/metric[3]:.3f} --- test acc {test_acc:.3f} --- lr {optimizer.state_dict()['param_groups'][0]['lr']} --- cost time {timer.sum()}")
#---------保存训练数据---------------
df = pd.DataFrame()
loss_list.append(metric[0] / metric[2])
train_acc_list.append(metric[1] / metric[3])
test_acc_list.append(test_acc)
epochs_list.append(epoch+1)
time_list.append(timer.sum())
lr_list.append(optimizer.state_dict()['param_groups'][0]['lr'])
df['epoch'] = epochs_list
df['loss'] = loss_list
df['train_acc'] = train_acc_list
df['test_acc'] = test_acc_list
df["lr"] = lr_list
df['time'] = time_list
df.to_excel("../blork_file/savefile/PSPNET.xlsx")
#----------------保存模型-------------------
if np.mod(epoch+1, 5) == 0:
torch.save(net, f'../blork_file/checkpoints/PSPNET{epoch+1}.pth')
# 保存下最后的model
torch.save(net, f'../blork_file/checkpoints/PSPNET.pth')训练结果
