网站推广教程分享我国哪些网站是做调查问卷的

网站推广教程分享,我国哪些网站是做调查问卷的,关键词推广优化排名如何,广州市南沙住房和建设局网站一、transforms的用法 transforms​ 是数据预处理与增强的核心工具#xff0c;主要用于将原始图像转换为模型可接受的格式#xff0c;并通过随机变换丰富数据集以提高模型泛化能力。 导入方式#xff1a; from torchvision import transforms主要用法#xff0c;按顺序 …一、transforms的用法transforms​ 是数据预处理与增强的核心工具主要用于将原始图像转换为模型可接受的格式并通过随机变换丰富数据集以提高模型泛化能力。导入方式fromtorchvisionimporttransforms主要用法按顺序transform_pipelinetransforms.Compose([transforms.Resize(256),# 调整图像大小至256x256保持宽高比transforms.CenterCrop(224),# 从中心裁剪224x224区域常用预训练模型输入尺寸transforms.RandomHorizontalFlip(p0.5),# 以50%概率水平翻转数据增强transforms.ToTensor(),# 将PIL图像转换为Tensor像素值缩放至[0,1]transforms.Normalize(# 标准化使用ImageNet均值/方差mean[0.485,0.456,0.406],# RGB通道均值std[0.229,0.224,0.225]# RGB通道标准差)])二、transform的使用将PIL图像转换成Tensor类型fromPILimportImagefromtorchvisionimporttransforms img_pathrdata/train/ants_image/0013035.jpgimgImage.open(img_path)tensor_transtransforms.ToTensor()tensor_imgtensor_trans(img)print(tensor_img.shape)#CHW通过tensor()类型的数据生成tensorboard图fromPILimportImagefromtorch.utils.tensorboardimportSummaryWriterfromtorchvisionimporttransforms img_pathrdata/train/ants_image/0013035.jpgimgImage.open(img_path)tensor_transtransforms.ToTensor()tensor_imgtensor_trans(img)# print(tensor_img.shape) #CHWwriterSummaryWriter(logs)writer.add_image(tensor_img,tensor_img,0)writer.close()Normalize()归一化使用fromPILimportImagefromtorch.utils.tensorboardimportSummaryWriterfromtorchvisionimporttransforms img_pathrdata/train/ants_image/0013035.jpgimgImage.open(img_path)tensor_transtransforms.ToTensor()tensor_imgtensor_trans(img)# print(tensor_img.shape) #CHWwriterSummaryWriter(logs)norm_transtransforms.Normalize([0.485,0.456,0.406],[0.5,0.5,0.5])norm_imgnorm_trans(tensor_img)writer.add_image(tensor_img,tensor_img,0)writer.add_image(norm_img,norm_img,1)writer.close()归一化后的图片和未归一化的图片Resize()调整大小的使用fromPILimportImagefromtorch.utils.tensorboardimportSummaryWriterfromtorchvisionimporttransforms img_pathrdata/train/ants_image/0013035.jpgimgImage.open(img_path)tensor_transtransforms.ToTensor()tensor_imgtensor_trans(img)# print(tensor_img.shape) #CHWwriterSummaryWriter(logs)norm_transtransforms.Normalize([0.485,0.456,0.406],[0.5,0.5,0.5])norm_imgnorm_trans(tensor_img)# print(img.size)resize_transtransforms.Resize((256,256))resize_imgresize_trans(tensor_img)writer.add_image(resize_img,resize_img,0)# print(resize_img.size)#Compose用法trans_resize_2transforms.Compose([transforms.Resize((512)),transforms.ToTensor()])img_resize_2trans_resize_2(img)writer.add_image(tensor_img,tensor_img,0)writer.add_image(norm_img,norm_img,1)writer.add_image(img_resize_2,img_resize_2,2)writer.close()