pytorch的语义分割------数据增⼴
语义分割的labe与分类问题不同，语义分割的label是⼀个mask，所以训练图像在做增⼴的时候，mask也要做相应的变换。
如：原来的图⽚和mask如下：
做旋转后：
做翻转后：
所以对于语义分割⽽⾔，其数据增⼴是连同mask⼀起做的。
---------------------------------------------------------------------------------------------------
以下提供⼀份代码，集成了 ansforms.functional ⾥的⽐较常⽤的数据增⼴⽅法(包括旋转，上下翻转，左右翻转，裁剪，调整对⽐度，调整饱和度，调整亮度，中⼼裁剪等)：
此代码⽤于在训练前通过对原来的数据做多种转换⽣产更多的训练数据。
此代码完全可以直接调⽤，⽤⼀个.py⽂件存放即可。
代码：
import random
import os
import numpy as np
from PIL import Image
from torchvision import transforms
ansforms.functional as tf
class Augmentation:
def __init__(self):
pass
def rotate(self,image,mask,angle=None):
if angle == None:
angle = _params([-180, 180]) # -180~180随机选⼀个⾓度旋转
if isinstance(angle,list):
angle = random.choice(angle)
image = ate(angle)
image = ate(angle)
mask = ate(angle)
image = tf.to_tensor(image)
mask = tf.to_tensor(mask)
return image, mask
def flip(self,image,mask): #⽔平翻转和垂直翻转
if random.random()>0.5:
image = tf.hflip(image)
mask = tf.hflip(mask)
if random.random()<0.5:
image = tf.vflip(image)
mask = tf.vflip(mask)
image = tf.to_tensor(image)
mask = tf.to_tensor(mask)
return image, mask
def randomResizeCrop(self,image,mask,scale=(0.3,1.0),ratio=(1,1)):#scale表⽰随机crop出来的图⽚会在的0.3倍⾄1倍之间，ratio表⽰长宽⽐
img = np.array(image)
h_image, w_image = img.shape
resize_size = h_image
i, j, h, w = _params(image, scale=scale, ratio=ratio)
image = tf.resized_crop(image, i, j, h, w, resize_size)
mask = tf.resized_crop(mask, i, j, h, w, resize_size)
image = tf.to_tensor(image)
mask = tf.to_tensor(mask)
return image, mask
def adjustContrast(self,image,mask):
factor = _params([0,10])  #这⾥调增⼴后的数据的对⽐度
image = tf.adjust_contrast(image,factor)
#mask = tf.adjust_contrast(mask,factor)
image = tf.to_tensor(image)
mask = tf.to_tensor(mask)
return image,mask
def adjustBrightness(self,image,mask):
factor = _params([1, 2])  #这⾥调增⼴后的数据亮度
image = tf.adjust_brightness(image, factor)
#mask = tf.adjust_contrast(mask, factor)
image = tf.to_tensor(image)
mask = tf.to_tensor(mask)
return image, mask
def centerCrop(self,image,mask,size=None): #中⼼裁剪
if size == None:size = image.size  #若不设定size，则是原图。
image = tf.center_crop(image,size)
mask = tf.center_crop(mask,size)
image = tf.to_tensor(image)
mask = tf.to_tensor(mask)
return image,mask
def adjustSaturation(self,image,mask): #调整饱和度
factor = _params([1, 2])  # 这⾥调增⼴后的数据亮度
image = tf.adjust_saturation(image, factor)
#mask = tf.adjust_saturation(mask, factor)
image = tf.to_tensor(image)
mask = tf.to_tensor(mask)
return image, mask
def augmentationData(image_path,mask_path,option=[1,2,3,4,5,6,7],save_dir=None):
'''
:param image_path: 图⽚的路径
:param mask_path: mask的路径
:param option: 需要哪种增⼴⽅式：1为旋转，2为翻转，3为随机裁剪并恢复原本⼤⼩，4为调整对⽐度，
5为中⼼裁剪(不恢复原本⼤⼩)，6为调整亮度,7为饱和度    :param save_dir: 增⼴后的数据存放的路径
'''
aug_image_savedDir = os.path.join(save_dir,'img')
aug_mask_savedDir = os.path.join(save_dir, 'mask')
if not ists(aug_image_savedDir):
os.makedirs(aug_image_savedDir)
os.makedirs(aug_image_savedDir)
print('create aug ')
if not ists(aug_mask_savedDir):
os.makedirs(aug_mask_savedDir)
print('create aug ')
aug = Augmentation()
res= os.walk(image_path)
images = []
masks = []
for root,dirs,files in res:
for f in files:
images.append(os.path.join(root,f))
res = os.walk(mask_path)
for root,dirs,files in res:
for f in files:
masks.append(os.path.join(root,f))
datas = list(zip(images,masks))
num = len(datas)
for (image_path,mask_path) in datas:
image = Image.open(image_path)
mask = Image.open(mask_path)
if 1 in option:
num+=1
image_tensor, mask_tensor = ate(image, mask)
image_rotate = transforms.ToPILImage()(image_tensor).save(os.path.join(save_dir, 'img', str(num) + '__rotate.jpg'))
mask_rotate = transforms.ToPILImage()(mask_tensor).save(os.path.join(save_dir, 'mask', str(num) + '_rotate_mask.jpg'))
if 2 in option:
num+=1
image_tensor, mask_tensor = aug.flip(image, mask)
image_filp = transforms.ToPILImage()(image_tensor).save(os.path.join(save_dir,'img',str(num)+'_filp.jpg'))
mask_filp = transforms.ToPILImage()(mask_tensor).save(os.path.join(save_dir,'mask',str(num)+'_filp_mask.jpg'))
if 3 in option:
num+=1
image_tensor, mask_tensor = aug.randomResizeCrop(image, mask)
image_ResizeCrop = transforms.ToPILImage()(image_tensor).save(os.path.join(save_dir, 'img', str(num) + '_ResizeCrop.jpg'))
mask_ResizeCrop = transforms.ToPILImage()(mask_tensor).save(os.path.join(save_dir, 'mask', str(num) + '_ResizeCrop_mask.jpg'))        if 4 in option:
num+=1
image_tensor, mask_tensor = aug.adjustContrast(image, mask)
image_Contrast = transforms.ToPILImage()(image_tensor).save(os.path.join(save_dir, 'img', str(num) + '_Contrast.jpg'))
mask_Contrast = transforms.ToPILImage()(mask_tensor).save(os.path.join(save_dir, 'mask', str(num) + '_Contrast_mask.jpg'))
if 5 in option:
num+=1
image_tensor, mask_tensor = Crop(image, mask)
image_centerCrop = transforms.ToPILImage()(image_tensor).save(os.path.join(save_dir, 'img', str(num) + '_centerCrop.jpg'))
resized
mask_centerCrop = transforms.ToPILImage()(mask_tensor).save(os.path.join(save_dir, 'mask', str(num) + '_centerCrop_mask.jpg'))        if 6 in option:
num+=1
image_tensor, mask_tensor = aug.adjustBrightness(image, mask)
image_Brightness = transforms.ToPILImage()(image_tensor).save(os.path.join(save_dir, 'img', str(num) + '_Brightness.jpg'))
mask_Brightness = transforms.ToPILImage()(mask_tensor).save(os.path.join(save_dir, 'mask', str(num) + '_Brightness.jpg'))
if 7 in option:
num+=1
image_tensor, mask_tensor = aug.adjustSaturation(image, mask)
image_Saturation = transforms.ToPILImage()(image_tensor).save(os.path.join(save_dir, 'img', str(num) + '_Saturation.jpg'))
mask_Saturation = transforms.ToPILImage()(mask_tensor).save(os.path.join(save_dir, 'mask', str(num) + '_Saturation.jpg')) augmentationData(r'E:\datasets\isbi\train\images',r'E:\datasets\isbi\train\label',save_dir=r'E:\代码\mytext\suanfa\aug')
运⾏效果：

版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。