网站中的二维码设计,网站建设费用计算,刷赞抖音推广网站,广州做网站建设对象分割是在图像中找到目标对象的边界的过程。单目标分割的重点是自动勾勒出图像中一个目标对象的边界。对象边界通常由二进制掩码定义。 通过二进制掩码#xff0c;可以在图像上覆盖轮廓以勾勒出对象边界。例如以下图片描绘了胎儿的超声图像、胎儿头部的二进制掩码以及覆盖在… 对象分割是在图像中找到目标对象的边界的过程。单目标分割的重点是自动勾勒出图像中一个目标对象的边界。对象边界通常由二进制掩码定义。 通过二进制掩码可以在图像上覆盖轮廓以勾勒出对象边界。例如以下图片描绘了胎儿的超声图像、胎儿头部的二进制掩码以及覆盖在超声图像上的胎儿头部的图像分割 目录
准备数据集
创建自定义数据集
划分数据集
创建数据加载器
搭建模型
定义损失函数
定义优化器
训练和评估模型 准备数据集
使用胎儿头围数据集Automated measurement of fetal head circumference在怀孕期间超声成像用于测量胎儿头围监测胎儿的生长。数据集包含标准平面的二维2D超声图像。Automated measurement of fetal head circumferenceFor more information about this dataset go to: https://hc18.grand-challenge.org/https://zenodo.org/record/1322001#.XcX1jk9KhhE
import os
path2train./data/training_set/imgsList[pp for pp in os.listdir(path2train) if Annotation not in pp]
anntsList[pp for pp in os.listdir(path2train) if Annotation in pp]
print(number of images:, len(imgsList))
print(number of annotations:, len(anntsList))import numpy as np
np.random.seed(2024)
rndImgsnp.random.choice(imgsList,4)
rndImgsimport matplotlib.pylab as plt
from PIL import Image
from scipy import ndimage as ndi
from skimage.segmentation import mark_boundaries
from torchvision.transforms.functional import to_tensor, to_pil_image
import torchdef show_img_mask(img, mask):if torch.is_tensor(img):imgto_pil_image(img)maskto_pil_image(mask)img_maskmark_boundaries(np.array(img), np.array(mask),outline_color(0,1,0),color(0,1,0))plt.imshow(img_mask)
for fn in rndImgs:path2img os.path.join(path2train, fn)path2annt path2img.replace(.png, _Annotation.png)img Image.open(path2img)annt_edges Image.open(path2annt)mask ndi.binary_fill_holes(annt_edges) plt.figure()plt.subplot(1, 3, 1) plt.imshow(img, cmapgray)plt.subplot(1, 3, 2) plt.imshow(mask, cmapgray)plt.subplot(1, 3, 3) show_img_mask(img, mask) plt.figure()
plt.subplot(1, 3, 1)
plt.imshow(img, cmapgray)
plt.axis(off)plt.subplot(1, 3, 2)
plt.imshow(mask, cmapgray)
plt.axis(off) plt.subplot(1, 3, 3)
show_img_mask(img, mask)
plt.axis(off)# conda install conda-forge/label/cf202003::albumentations
from albumentations import (HorizontalFlip,VerticalFlip, Compose,Resize,
)h,w128,192
transform_train Compose([ Resize(h,w), HorizontalFlip(p0.5), VerticalFlip(p0.5), ])transform_val Resize(h,w) 创建自定义数据集
from torch.utils.data import Dataset
from PIL import Image
from torchvision.transforms.functional import to_tensor, to_pil_imageclass fetal_dataset(Dataset):def __init__(self, path2data, transformNone): imgsList[pp for pp in os.listdir(path2data) if Annotation not in pp]anntsList[pp for pp in os.listdir(path2train) if Annotation in pp]self.path2imgs [os.path.join(path2data, fn) for fn in imgsList] self.path2annts [p2i.replace(.png, _Annotation.png) for p2i in self.path2imgs]self.transform transformdef __len__(self):return len(self.path2imgs)def __getitem__(self, idx):path2img self.path2imgs[idx]image Image.open(path2img)path2annt self.path2annts[idx]annt_edges Image.open(path2annt)mask ndi.binary_fill_holes(annt_edges) image np.array(image)maskmask.astype(uint8) if self.transform:augmented self.transform(imageimage, maskmask)image augmented[image]mask augmented[mask] image to_tensor(image) mask255*to_tensor(mask) return image, maskfetal_ds1fetal_dataset(path2train, transformtransform_train)
fetal_ds2fetal_dataset(path2train, transformtransform_val)
img,maskfetal_ds1[0]
print(img.shape, img.type(),torch.max(img))
print(mask.shape, mask.type(),torch.max(mask))show_img_mask(img, mask) 划分数据集
按照82的比例划分训练数据集和验证数据集
from sklearn.model_selection import ShuffleSplitsss ShuffleSplit(n_splits1, test_size0.2, random_state0)
indicesrange(len(fetal_ds1))
for train_index, val_index in sss.split(indices):print(len(train_index))print(-*10)print(len(val_index)) from torch.utils.data import Subsettrain_dsSubset(fetal_ds1,train_index)
print(len(train_ds))
val_dsSubset(fetal_ds2,val_index)
print(len(val_ds)) 展示训练数据集示例图像
plt.figure(figsize(5,5))
for img,mask in train_ds:show_img_mask(img,mask)break 展示验证数据集示例图像
plt.figure(figsize(5,5))
for img,mask in val_ds:show_img_mask(img,mask)break 创建数据加载器
from torch.utils.data import DataLoader
train_dl DataLoader(train_ds, batch_size8, shuffleTrue)
val_dl DataLoader(val_ds, batch_size16, shuffleFalse) for img_b, mask_b in train_dl:print(img_b.shape,img_b.dtype)print(mask_b.shape, mask_b.dtype)breakfor img_b, mask_b in val_dl:print(img_b.shape,img_b.dtype)print(mask_b.shape, mask_b.dtype)breaktorch.max(img_b) 搭建模型
基于编码器-解码器模型encoder–decoder model搭建分割任务模型 import torch.nn as nn
import torch.nn.functional as Fclass SegNet(nn.Module):def __init__(self, params):super(SegNet, self).__init__()C_in, H_in, W_inparams[input_shape]init_fparams[initial_filters] num_outputsparams[num_outputs] self.conv1 nn.Conv2d(C_in, init_f, kernel_size3,stride1,padding1)self.conv2 nn.Conv2d(init_f, 2*init_f, kernel_size3,stride1,padding1)self.conv3 nn.Conv2d(2*init_f, 4*init_f, kernel_size3,padding1)self.conv4 nn.Conv2d(4*init_f, 8*init_f, kernel_size3,padding1)self.conv5 nn.Conv2d(8*init_f, 16*init_f, kernel_size3,padding1)self.upsample nn.Upsample(scale_factor2, modebilinear, align_cornersTrue)self.conv_up1 nn.Conv2d(16*init_f, 8*init_f, kernel_size3,padding1)self.conv_up2 nn.Conv2d(8*init_f, 4*init_f, kernel_size3,padding1)self.conv_up3 nn.Conv2d(4*init_f, 2*init_f, kernel_size3,padding1)self.conv_up4 nn.Conv2d(2*init_f, init_f, kernel_size3,padding1)self.conv_out nn.Conv2d(init_f, num_outputs , kernel_size3,padding1) def forward(self, x):x F.relu(self.conv1(x))x F.max_pool2d(x, 2, 2)x F.relu(self.conv2(x))x F.max_pool2d(x, 2, 2)x F.relu(self.conv3(x))x F.max_pool2d(x, 2, 2)x F.relu(self.conv4(x))x F.max_pool2d(x, 2, 2)x F.relu(self.conv5(x))xself.upsample(x)x F.relu(self.conv_up1(x))xself.upsample(x)x F.relu(self.conv_up2(x))xself.upsample(x)x F.relu(self.conv_up3(x))xself.upsample(x)x F.relu(self.conv_up4(x))x self.conv_out(x)return x params_model{input_shape: (1,h,w),initial_filters: 16, num_outputs: 1,}model SegNet(params_model)import torch
device torch.device(cuda if torch.cuda.is_available() else cpu)
modelmodel.to(device)
打印模型结构
print(model) 获取模型摘要
from torchsummary import summary
summary(model, input_size(1, h, w)) 定义损失函数
def dice_loss(pred, target, smooth 1e-5):intersection (pred * target).sum(dim(2,3))union pred.sum(dim(2,3)) target.sum(dim(2,3)) dice 2.0 * (intersection smooth) / (union smooth) loss 1.0 - dicereturn loss.sum(), dice.sum()import torch.nn.functional as Fdef loss_func(pred, target):bce F.binary_cross_entropy_with_logits(pred, target, reductionsum)pred torch.sigmoid(pred)dlv, _ dice_loss(pred, target)loss bce dlvreturn lossfor img_v,mask_v in val_dl:mask_v mask_v[8:]breakfor img_t,mask_t in train_dl:breakprint(dice_loss(mask_v,mask_v))
loss_func(mask_v,torch.zeros_like(mask_v))import torchvisiondef metrics_batch(pred, target):pred torch.sigmoid(pred)_, metricdice_loss(pred, target)return metricdef loss_batch(loss_func, output, target, optNone): loss loss_func(output, target)with torch.no_grad():pred torch.sigmoid(output)_, metric_bdice_loss(pred, target)if opt is not None:opt.zero_grad()loss.backward()opt.step()return loss.item(), metric_b 定义优化器
from torch import optim
opt optim.Adam(model.parameters(), lr3e-4)from torch.optim.lr_scheduler import ReduceLROnPlateau
lr_scheduler ReduceLROnPlateau(opt, modemin,factor0.5, patience20,verbose1)def get_lr(opt):for param_group in opt.param_groups:return param_group[lr]current_lrget_lr(opt)
print(current lr{}.format(current_lr)) 训练和评估模型
def loss_epoch(model,loss_func,dataset_dl,sanity_checkFalse,optNone):running_loss0.0running_metric0.0len_datalen(dataset_dl.dataset)for xb, yb in dataset_dl:xbxb.to(device)ybyb.to(device)outputmodel(xb)loss_b, metric_bloss_batch(loss_func, output, yb, opt)running_loss loss_bif metric_b is not None:running_metricmetric_bif sanity_check is True:breaklossrunning_loss/float(len_data)metricrunning_metric/float(len_data)return loss, metric
import copy
def train_val(model, params):num_epochsparams[num_epochs]loss_funcparams[loss_func]optparams[optimizer]train_dlparams[train_dl]val_dlparams[val_dl]sanity_checkparams[sanity_check]lr_schedulerparams[lr_scheduler]path2weightsparams[path2weights]loss_history{train: [],val: []}metric_history{train: [],val: []} best_model_wts copy.deepcopy(model.state_dict())best_lossfloat(inf) for epoch in range(num_epochs):current_lrget_lr(opt)print(Epoch {}/{}, current lr{}.format(epoch, num_epochs - 1, current_lr)) model.train()train_loss, train_metricloss_epoch(model,loss_func,train_dl,sanity_check,opt)loss_history[train].append(train_loss)metric_history[train].append(train_metric)model.eval()with torch.no_grad():val_loss, val_metricloss_epoch(model,loss_func,val_dl,sanity_check)loss_history[val].append(val_loss)metric_history[val].append(val_metric) if val_loss best_loss:best_loss val_lossbest_model_wts copy.deepcopy(model.state_dict())torch.save(model.state_dict(), path2weights)print(Copied best model weights!)lr_scheduler.step(val_loss)if current_lr ! get_lr(opt):print(Loading best model weights!)model.load_state_dict(best_model_wts) print(train loss: %.6f, dice: %.2f %(train_loss,100*train_metric))print(val loss: %.6f, dice: %.2f %(val_loss,100*val_metric))print(-*10) model.load_state_dict(best_model_wts)return model, loss_history, metric_history opt optim.Adam(model.parameters(), lr3e-4)# 定义学习率调度器当验证集上的损失不再下降时将学习率降低为原来的0.5倍等待20个epoch后再次降低学习率
lr_scheduler ReduceLROnPlateau(opt, modemin,factor0.5, patience20,verbose1)path2models ./models/# 判断path2models路径是否存在如果不存在则创建该路径
if not os.path.exists(path2models):os.mkdir(path2models)params_train{num_epochs: 100,optimizer: opt,loss_func: loss_func,train_dl: train_dl,val_dl: val_dl,sanity_check: False,lr_scheduler: lr_scheduler,path2weights: path2modelsweights.pt,
}model,loss_hist,metric_histtrain_val(model,params_train) 打印训练验证损失
num_epochsparams_train[num_epochs]plt.title(Train-Val Loss)
plt.plot(range(1,num_epochs1),loss_hist[train],labeltrain)
plt.plot(range(1,num_epochs1),loss_hist[val],labelval)
plt.ylabel(Loss)
plt.xlabel(Training Epochs)
plt.legend()
plt.show() 打印训练验证精度
# plot accuracy progress
plt.title(Train-Val Accuracy)
plt.plot(range(1,num_epochs1),metric_hist[train],labeltrain)
plt.plot(range(1,num_epochs1),metric_hist[val],labelval)
plt.ylabel(Accuracy)
plt.xlabel(Training Epochs)
plt.legend()
plt.show()
