数据变换Transforms
- 环境准备
# 实验环境已经预装了mindspore==2.2.14,如需更换mindspore版本,可更改下面mindspore的版本号
!pip uninstall mindspore -y
!pip install -i https://pypi.mirrors.ustc.edu.cn/simple mindspore==2.2.14
import numpy as np
from PIL import Image
from download import download
from mindspore.dataset import transforms, vision, text
from mindspore.dataset import GeneratorDataset, MnistDataset
Common Transforms
Compose
Compose
接收一个数据增强操作序列,然后将其组合成单个数据增强操作。基于Mnist数据集呈现Transforms的应用效果#下载开源数据集 url = "https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/" \ "notebook/datasets/MNIST_Data.zip" path = download(url, "./", kind="zip", replace=True) train_dataset = MnistDataset('MNIST_Data/train') image, label = next(train_dataset.create_tuple_iterator()) print(image.shape) #output (28, 28, 1) composed = transforms.Compose([ vision.Rescale(1.0 / 255.0, 0), vision.Normalize(mean=(0.1307,), std=(0.3081,)), vision.HWC2CHW() ]) train_dataset = train_dataset.map(composed, 'image') image, label = next(train_dataset.create_tuple_iterator()) print(image.shape) #output(1, 28, 28)
Vision Transforms
- Rescale
- 调整图像像素值的大小,包括两个参数:
- rescale: 缩放因子
- shift: 平移因子
- 输出像素值为 o u t p u t i = i n p u t i ∗ r e s c a l e + s h i f t output_i=input_i * rescale + shift outputi=inputi∗rescale+shift
random_np = np.random.randint(0, 255, (48, 48), np.uint8) random_image = Image.fromarray(random_np) rescale = vision.Rescale(1.0 / 255.0, 0) rescaled_image = rescale(random_image)
- 调整图像像素值的大小,包括两个参数:
- Normalize
- 对输入图像的归一化,包括三个参数:
- mean: 图像每个通道的均值。
- std: 图像每个通道的标准差。
- is_hwc: bool值,输入图像的格式。True为(height, width, channel), False为(channel, height, width)。
- 图像通道根据
mean
和std
进行调整,计算公式 o u t p u t c = i n p u t c − m e a n c s t d c , c output_c = \frac{input_c - mean_c}{std_c}, c outputc=stdcinputc−meanc,c 代表通道索引。
normalize = vision.Normalize(mean=(0.1307,), std=(0.3081,)) normalized_image = normalize(rescaled_image)
- 对输入图像的归一化,包括三个参数:
- HWC2CHW
HWC2CHW
变换用于转换图像格式。hwc_image = np.expand_dims(normalized_image, -1) hwc2chw = vision.HWC2CHW() chw_image = hwc2chw(hwc_image) print(hwc_image.shape, chw_image.shape) #Output (48, 48, 1)(1, 48, 48)
Text Transforms
文本数据需要有分词(Tokenize)、构建词表、Token转Index等操作。
text = ['Welcome to Beijing']
text_dataset = GeneratorDataset(texts, 'text')
- PythonTokenizer 分词器
def my_tokenizer(content): return content.split() test_dataset = test_dataset.map(text.PythonTokenizer(my_tokenizer)) print(next(test_dataset.create_tuple_iterator()) #Output [Tensor(shape=[3], dtype=String, value= ['Welcome', 'to', 'Beijing'])]
- Lookup
Lookup
为词表映射变换,用来将Token转化为index。- 前置条件: 需要构造词表
vocab = text.Vocab.from_dataset(test_dataset) print(vocab.vocab()) #Output {'to': 2, 'Welcome': 1, 'Beijing': 0} test_dataset = test_dataset.map(text.Lookup(vocab)) print(next(test_dataset.create_tuple_iterator())) #Output [Tensor(shape=[3], dtype=Int32, value= [1, 2, 0])]
- 前置条件: 需要构造词表
- Lambda Transforms
- 加载任意定义的Lambda函数
test_dataset = GeneratorDataset([1, 2, 3], 'data', shuffle=False) test_dataset = test_dataset.map(lambda x: x * 2) def func(x): return x * x + 2 test_dataset = test_dataset.map(lambda x: func(x))
- 加载任意定义的Lambda函数