AugLy图像增强方法

Livingbody
1,540 阅读3分钟

「这是我参与2022首次更文挑战的第6天，活动详情查看：2022首次更文挑战」。

0 前言

Augly是由facebook ai开源的一款用于数据增强的python库，它支持音频、图像、视频、文本4种形式的数据，提供了超过100种数据增强功能，可以直接对输入的内容进行多种处理。对于图像而言，提供了裁剪、选择图像、图像加文字、增加饱和度、亮度调整等100多种数据增强的方式，pip安装，开箱即用，很方便我们用来对图像进行处理。可以说功能非常的强大了，不仅可以用来离线数据增强，还可以在torch的训练中进行在线的数据增强，具体怎么使用，可以参考官方给出的例子。

目前本文给出以下增强的方法

  • 图像模糊
  • 改变图像亮度
  • 改变图像高宽比
  • 颜色晃动
  • 对比度增强
  • 按比例裁剪
  • 改变图像质量
  • 转灰度
  • 水平翻转
  • 图片上方创建文字
  • 改变图像的不透明度
  • 图片添加表情
  • 两幅图像重叠
  • 网页覆盖
  • 图像中间加横条
  • 图片添加文字
  • 图像边界填充
  • 填充短边，成为正方形
  • 透视变换
  • 马赛克
  • 随机噪声
  • 图像旋转
  • 改变图像饱和度
  • 改变图像分辨率
  • 锐化
  • 随机像素比任意变化
  • 垂直翻转

1 Augly安装

!pip install augly

Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Requirement already satisfied: augly in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (0.1.9)
Requirement already satisfied: Pillow>=8.2.0 in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (from augly) (8.3.2)
Requirement already satisfied: numpy>=1.19.5 in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (from augly) (1.21.2)
Requirement already satisfied: nlpaug==1.1.3 in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (from augly) (1.1.3)
Requirement already satisfied: python-magic>=0.4.22 in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (from augly) (0.4.24)
Requirement already satisfied: regex>=2021.4.4 in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (from augly) (2021.9.30)
Requirement already satisfied: iopath>=0.1.8 in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (from augly) (0.1.9)
Requirement already satisfied: portalocker in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (from iopath>=0.1.8->augly) (2.3.2)
Requirement already satisfied: tqdm in /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages (from iopath>=0.1.8->augly) (4.36.1)

2 常用的图像增强接口

import augly.image as imaugs
import PIL.Image as Image

image_path = "day.jpeg"

aug_image = Image.open(image_path)
aug_image.show()

output_4_0.png

# 图像模糊
output_path = "work/blur.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.blur(aug_image,output_path=output_path)

aug_image.show()

output_5_0.png

# 改变图像亮度
output_path = "work/brightness.jpeg"             # 输出图片路径

aug_image = Image.open(image_path)
aug_image = imaugs.brightness(aug_image,factor=1.5,output_path=output_path)     # 修改factor参数调整亮度

aug_image.show()

output_6_0.png

# 改变图像高宽比
output_path = "work/change_aspect_ratio.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.change_aspect_ratio(aug_image,ratio=2,output_path=output_path)

aug_image.show()

output_7_0.png

# 颜色晃动
output_path = "work/color_jitter.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.color_jitter(aug_image,brightness_factor=1.0,contrast_factor=1.0,saturation_factor=1.5,output_path=output_path)

aug_image.show()

output_8_0.png

# 对比度增强
output_path = "work/contrast.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.contrast(aug_image,factor=1.5,output_path=output_path)

aug_image.show()

output_9_0.png

# 按比例裁剪
output_path = "work/crop.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.crop(aug_image,x1=0.1,y1=0.1,x2=0.9,y2=0.9,output_path=output_path)

aug_image.show()

output_10_0.png

# 改变图像质量
output_path = "work/encoding_quality.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.encoding_quality(aug_image,output_path=output_path)

aug_image.show()

output_11_0.png

# 转灰度
output_path = "work/grayscale.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.grayscale(aug_image,output_path=output_path)

aug_image.show()

output_12_0.png

# 水平翻转
output_path = "work/hflip.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.hflip(aug_image,output_path=output_path)

aug_image.show()

output_13_0.png

# 图片上方创建文字
output_path = "work/meme_format.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.meme_format(aug_image,text="hello world",caption_height=50,output_path=output_path)

aug_image.show()

output_14_0.png

# 改变图像的不透明度
output_path = "work/opacity.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.opacity(aug_image,level=0.5,output_path=output_path)

aug_image.show()

output_15_0.png

# 图片添加表情
output_path = "work/overlay_emoji.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.overlay_emoji(aug_image,output_path=output_path)

aug_image.show()

output_16_0.png

# 两幅图像重叠
image_path2 = "day1.jpeg"
output_path = "work/overlay_image.jpeg"

aug_image1 = Image.open(image_path)
aug_image2 = Image.open(image_path2)

aug_image = imaugs.overlay_image(aug_image1,image_path2,output_path=output_path)

aug_image.show()

output_17_0.png

# 网页覆盖
output_path = "work/overlay_onto_screenshot.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.overlay_onto_screenshot(aug_image,output_path=output_path)

aug_image.show()

output_18_0.png

# 图像中间加横条
output_path = "work/overlay_stripes.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.overlay_stripes(aug_image,line_width=0.1,output_path=output_path)

aug_image.show()

output_19_0.png

# 图片添加文字
output_path = "work/overlay_text.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.overlay_text(aug_image,output_path=output_path)

aug_image.show()

output_20_0.png

# 图像边界填充
output_path = "work/pad.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.pad(aug_image,w_factor=0.25,h_factor=0.25,output_path=output_path)

aug_image.show()

output_21_0.png

# 填充短边，成为正方形
output_path = "work/pad_square.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.pad_square(aug_image,color=(255,255,255),output_path=output_path)

aug_image.show()

output_22_0.png

# 透视变换
output_path = "work/perspective_transform.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.perspective_transform(aug_image,output_path=output_path)

aug_image.show()

output_23_0.png

# 马赛克
output_path = "work/pixelization.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.pixelization(aug_image,ratio=0.5,output_path=output_path)

aug_image.show()

output_24_0.png

# 随机噪声
output_path = "work/random_noise.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.random_noise(aug_image,output_path=output_path)

aug_image.show()

output_25_0.png

# 图像旋转
output_path = "work/rotate.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.rotate(aug_image,degrees=10,output_path=output_path)

aug_image.show()

output_26_0.png

# 改变图像饱和度
output_path = "work/saturation.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.saturation(aug_image,factor=1.5,output_path=output_path)

aug_image.show()

output_27_0.png

# 改变图像分辨率
output_path = "work/scale.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.scale(aug_image,factor=0.8,output_path=output_path)

aug_image.show()

output_28_0.png

# 锐化
output_path = "work/sharpen.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.sharpen(aug_image,factor=1.5,output_path=output_path)

aug_image.show()

output_29_0.png

# 随机像素比任意变化
output_path = "work/shuffle_pixels.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.shuffle_pixels(aug_image,factor=0.1,output_path=output_path)

aug_image.show()

output_30_0.png

# 垂直翻转
output_path = "work/vflip.jpeg"

aug_image = Image.open(image_path)
aug_image = imaugs.vflip(aug_image,output_path=output_path)

aug_image.show()

output_31_0.png

avatar
文章
阅读
粉丝