weixin_38907184的博客

# -*- coding:utf-8 -*-import os.pathfrom collections import defaultdict, OrderedDictimport jsonimport numpy as npimport cv2import json# /8Tdata/gaofei/data/tomo/tianjin/b1_106/image2/# /8Tdata/gaofei/data/tomo/tianjin/b2_400/image2/#json文件with.

使用python的opencv将图像分割后的mask轮廓叠加在原图上

# encoding:utf-8import json,xlwtdef writeM(file): # 0 1 title = ["patientID","nodeID ",# 2 3 4 5 6 7 "2D-是否可见", "2D-形状","2D-密度","2D-边缘","2D-伴随钙化",.

import osfrom collections import defaultdict, OrderedDictimport jsonimport numpy as npimport cv2import globclass NpEncoder(json.JSONEncoder): def default(self, obj): if isinstance(obj, np.integer): return int(obj) el.

import osimport cv2import randomimport shutilimport numpy as npdef get_data_crop(k): # 获取所有病人序列名称与所有训练文件名称 name_list, file_list = [], [] for root, dirs, files in os.walk('./seg'): for f in files: if f.split('-')[0] n.

import numpy as npimport cv2import osfrom PIL import Imageimport matplotlib.pyplot as pltimbgr = cv2.imread('./pre/1464455/068.png')imgray = cv2.imread('./data/068.png',0)# 以灰度图读取mask图片，那么不同物体对应的像素值是不同的，提取统一像素值的所有位置#img = Image.open('./data/2.

import nibabel as nibdef save_array_as_nii_volume(data, filename, reference_name=None): # img = sitk.GetImageFromArray(data) # if (reference_name is not None): # img_ref = sitk.ReadImage(reference_name) # img.CopyInformation(img_.

import SimpleITK as sitkimport globimport numpy as npfrom PIL import Imageimport cv2def save_array_as_nii_volume(data, filename, reference_name=None): img = sitk.GetImageFromArray(data) if (reference_name is not None): img_ref = si.

import SimpleITK as sitkimport globimport numpy as npfrom PIL import Imageimport cv2def save_array_as_nii_volume(data, filename, reference_name=None): img = sitk.GetImageFromArray(data) if (reference_name is not None): img_ref = si.

import osimport SimpleITK as sitkdef readdcm(filepath): #最终发现是输入路径有中文，SimpleITK.ImageSeriesReader.GetGDCMSeriesFileNames(path)中path不能含有中文，否则无法读取。 series_id = sitk.ImageSeriesReader.GetGDCMSeriesIDs(filepath) series_file_names = sitk.ImageSe.

import numpy as npdef NMS(nums, threshold): x1 = nums[:, 0] y1 = nums[:, 1] x2 = nums[:, 2] y2 = nums[:, 3] area = (x2 - x1) * (y2 - y1) score = nums[:, 4] score = np.argsort(score)[::-1] # 函数是将x中的元素从小到大排列，提取其对应的index(索引)，.

def dice_coef(output, target):#output为预测结果 target为真实结果 smooth = 1e-5 #防止0除 if torch.is_tensor(output): output = torch.sigmoid(output).data.cpu().numpy() if torch.is_tensor(target): target = target.data.cpu().numpy() inter.

通过参考经验丰富的放射科医生的放射学报告，并遵循以下标准将肝肿物分为5类:手术或活检后的组织病理学评估(A-C类);影像学标准(A类，早期增强和延迟洗脱;C类[高血管病变]，延迟相图像早期增强等衰减;C类[低血管病变]，动脉期低衰减，延迟期等衰减或低衰减;D类[血管瘤]，动脉期周围结节强化，延迟期呈向心型充血强化，或动脉期均匀强化[接近主动脉衰减]，延迟期高衰减;E类[囊肿]，水衰减，无明显对比度增强);分别用未增强、动脉期和延迟期CT图像进行训练像素大小：70*70CNN由几个层组成(6个卷积

train_curve = list()def train_net(net, device, data_path, epochs=100, batch_size=4, lr=0.01): # 加载训练集 isbi_dataset = ISBI_Loader(data_path) train_loader = torch.utils.data.DataLoader(dataset=isbi_dataset, .

import cv2import globimport osimport numpy as npfrom medpy import metricdef calculate_metric_percase(pred, gt): # dice = metric.binary.dc(pred, gt) # jc = metric.binary.jc(pred, gt) hd = metric.binary.hd95(pred, gt) asd = metric.bina.

机构：约翰·霍普金斯大学/ 中国电子科技大学/ 斯坦福德大学 / 华东师范大学相关工作：CNN / Transformpatch size 消融实验：8 / 16 / 32

机构：香港大学 / 厦门大学相关工作（较新）：主干基于卷积/ 主干基于transform结构：流程图总揽 / 各个模块作介绍实验：有无预训练模型

数据集：ACDC/Synapse指标：dice / Hd相关工作：CNN /Transformer /Attention消融实验：上下采样/ 跳跃连接 / 图片输入大小

#测试预测与真值文件夹PSNR和SSIM#注意：需要两个文件夹有相同的顺序import cv2import globimport osfrom util.metrics import PSNRfrom util.metrics import SSIMpsnr_total = 0ssim_total = 0# predlist=glob.glob(os.path.join('../Datasets/GoPro/test/results/GoPro', "*"))#预测文件夹路径# g.

import osimport cv2import randomimport shutilimport numpy as npdef get_data_crop(k): # 获取所有病人序列名称与所有训练文件名称 name_list, file_list = [], [] for root, dirs, files in os.walk('./img'): for f in files: if f.split('-')[-2] .

'cv2'import cv2import osimport glob path = r'C:\Users\87419\Desktop\cg1\img\*.jpg'for i in glob.glob(path): im1 = cv2.imread(i) im2 = cv2.resize(im1,(256,256)) cv2.imwrite(os.path.join(r'C:\Users\87419\Desktop\cg1\resize',os.path.basenam.

输出中间训练过程图片 if step % 8000 == 0: s_img = image[0].permute(1, 2, 0).cpu().numpy().astype(np.int) s_lab = np.array(label[0].permute(1, 2, 0).cpu().numpy()*255, np.int) s_pre = np.array(torch.sigmoid(pred)[0].

https://blog.youkuaiyun.com/Accelerato/article/details/105627078

第一周 Alex net2020-06-15 00:00 至 2021-12-31 00:00作业内容：1.文字回答：ImageNet数据集与ILSVRC之间的关系是什么？ImageNet数据是ILSVRC比赛使用的数据集2.文字回答：AlexNet训练过程的数据增强中，针对位置，采用什么方法将一张图片有可能变为2048张不一样的图片（1个像素值不一样，也叫不一样）？训练阶段：1.图片统一缩放至256*256 2.随机位置裁剪出224*224区域 3.随机进行水平翻转测试阶...