深度学习《Data Augmentation》

1、数据增强 (tensorflow实现)

import tensorflow as tf
from matplotlib import pyplot as plt
import cv2 as cv
import numpy as np
import random

def get_img_from_path(path):
    ran = random.randint(0,10)
    print(ran)
    orgimg = tf.io.read_file(path)
    orgimg = tf.image.decode_jpeg(orgimg, channels=3)
    orgimg = tf.image.resize(orgimg, [168, 250])
    orgimg = tf.cast(orgimg, dtype="uint8")
    if ran == 0: 
        augimg = tf.image.random_flip_left_right(orgimg)         #'''随机左右翻转'''
    elif ran == 1: 
        augimg = tf.image.random_flip_up_down(orgimg)            #'''随机上下翻转'''
    elif ran == 2: 
        augimg = tf.image.random_hue(orgimg, max_delta=0.4)      #'''随机色度'''
    elif ran == 3: 
        augimg = tf.image.random_contrast(orgimg, lower=0.5, upper=1.8)    #'''随机对比度'''
    elif ran == 4: 
        augimg = tf.image.random_brightness(orgimg, max_delta=0.2)         #'''随机亮度'''
    elif ran == 5: 
        augimg = tf.image.random_saturation(orgimg, lower=0.2, upper=1.8)  #'''随机饱和度'''
    elif ran == 6: 
    	#'''随机图片质量'''
        augimg = tf.image.random_jpeg_quality(orgimg, min_jpeg_quality = 5, max_jpeg_quality = 100 ) 
    else:
        augimg = orgimg
    return orgimg, augimg

while 1:
    orgimg, augimg = get_img_from_path("test.jpg")
    cv.imshow("orgimg", orgimg.numpy())
    cv.imshow("augimg", augimg.numpy())
    key = cv.waitKey(0)
    if key == 27 : # ESC退出
        break
cv.destroyAllWindows()

ran == 0:(左边原图，右边数据增强后的图)

ran == 1:(左边原图，右边数据增强后的图)

ran == 2:(左边原图，右边数据增强后的图)

ran == 3:(左边原图，右边数据增强后的图)

ran == 4:(左边原图，右边数据增强后的图)

ran == 5:(左边原图，右边数据增强后的图)

ran == 6:(左边原图，右边数据增强后的图)