语义分割 - FCN ： 数据预处理/数据增强 算法源码实例
base_size=520
crop_size=480
flip_prob=0.5if train_val == 'train':self.transforms = transforms.Compose([transforms.RandomResize(int(base_size*0.5), int(base_size*2)),transforms.RandomHorizontalFlip(flip_prob),transforms.RandomCrop(crop_size),transforms.ToTensor(),transforms.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225))])
else:self.transforms = transforms.Compose([transforms.RandomResize(base_size, base_size),transforms.ToTensor(),transforms.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225))])
import numpy as np
import random
import torch
from torchvision import transforms
from torchvision.transforms import functional as Fdef pad_if_smaller(img, size, fill=0):# 如果图像最小边长小于给定size，则用数值fill进行paddingmin_size = min(img.size)if min_size < size:ow, oh = img.sizepadh = size - oh if oh < size else 0padw = size - ow if ow < size else 0img = F.pad(img, [0, 0, padw, padh], fill=fill)return imgclass Compose(object):def __init__(self, transforms):self.transforms = transformsdef __call__(self, image, target):for t in self.transforms:image, target = t(image, target)return image, targetclass RandomResize(object):def __init__(self, min_size, max_size=None):self.min_size = min_sizeif max_size is None:max_size = min_sizeself.max_size = max_sizedef __call__(self, image, target):size = random.randint(self.min_size, self.max_size)# 将image 和 target 的短边缩放到size大小image = F.resize(image, [size])target = F.resize(target, [size], interpolation=transforms.InterpolationMode.NEAREST)return image, targetclass RandomHorizontalFlip(object):def __init__(self, flip_prob):self.flip_prob = flip_probdef __call__(self, image, target):if random.random() < self.flip_prob:image = F.hflip(image)target = F.hflip(target)return image, targetclass RandomCrop(object):def __init__(self, size):self.size = sizedef __call__(self, image, target):image = pad_if_smaller(image, self.size)target = pad_if_smaller(target, self.size, fill=255)crop_params = transforms.RandomCrop.get_params(image, (self.size, self.size))image = F.crop(image, *crop_params)target = F.crop(target, *crop_params)return image, targetclass CenterCrop(object):def __init__(self, size):self.size = sizedef __call__(self, image, target):image = F.center_crop(image, self.size)target = F.center_crop(target, self.size)return image, targetclass ToTensor(object):def __call__(self, image, target):image = F.to_tensor(image)target = torch.as_tensor(np.array(target), dtype=torch.int64)return image, targetclass Normalize(object):def __init__(self, mean, std):self.mean = meanself.std = stddef __call__(self, image, target):image = F.normalize(image, mean=self.mean, std=self.std)return image, target
