tensorflow 数据增强
目录
说明：在以下方法中，可对一批图像进行处理，此处为了方便只用了一张图片作为例子来展示。
1.按角度旋转图片
import cv2
import matplotlib.pyplot as plt
import tensorflow as tf
import numpy as np
from math import pi
'''按角度旋转图片'''
def rotate_images(X_imgs, start_angle, end_angle, n_images,image_size_w,image_size_h):
X_rotate = []
iterate_at = (end_angle - start_angle) / (n_images - 1)
tf.reset_default_graph()
X = tf.placeholder(tf.float32, shape=(None, image_size_w,image_size_h, 3))
radian = tf.placeholder(tf.float32, shape=(len(X_imgs)))
tf_img = tf.contrib.image.rotate(X, radian)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for index in range(n_images):
degrees_angle = start_angle + index * iterate_at
radian_value = degrees_angle * pi / 180 # Convert to radian
radian_arr = [radian_value] * len(X_imgs)
rotated_imgs = sess.run(tf_img, feed_dict={X: X_imgs, radian: radian_arr})
X_rotate.extend(rotated_imgs)
X_rotate = np.array(X_rotate, dtype=np.float32)
return X_rotate
image = cv2.imread('plane.jpeg')
#image = tf.gfile.GFile('/home/ubuntu/images/output.jpg', 'wb'
b,g,r = cv2.split(image)
image = cv2.merge([r,g,b])
image = cv2.resize(image, (200,200),0,0, cv2.INTER_LINEAR) #图像缩放
image = np.multiply(image, 1.0 / 255.0)
list_image = []
list_image.append(image)
list_image = np.array(list_image)
#scaled_imgs = central_scale_images(list_image, [1.30, 1.50, 1.80])
scaled_imgs = rotate_images(list_image, -90, 90, 5,200,200)
#image = scaled_imgs[0]/255.
plt.subplot(331), plt.imshow(image), plt.title('', fontsize='medium')
plt.subplot(332), plt.imshow(scaled_imgs[0]), plt.title('',fontsize='small')
plt.subplot(333), plt.imshow(scaled_imgs[1]), plt.title('', fontsize='small')
plt.subplot(334), plt.imshow(scaled_imgs[2]), plt.title('', fontsize='small')
plt.subplot(335), plt.imshow(scaled_imgs[3]), plt.title('', fontsize='small')
plt.subplot(336), plt.imshow(scaled_imgs[4]), plt.title('', fontsize='small')
#plt.subplot(337), plt.imshow(scaled_imgs[5]), plt.title('', fontsize='small')
#plt.subplot(258), plt.imshow(scaled_imgs[6]), plt.title('Translate bottom 20 percent', fontsize='medium')
plt.show() 结果：
2. 按比例翻转
from math import ceil, floor
import numpy as np
import tensorflow as tf
import cv2
import matplotlib.pyplot as plt
def get_translate_parameters(index,image_size_w,image_size_h):
if index == 0: # Translate left 20 percent
offset = np.array([0.0, 0.3], dtype=np.float32)
size = np.array([image_size_w, ceil(0.7 * image_size_h)], dtype=np.int32)
w_start = 0
w_end = int(ceil(0.7 * image_size_w))
h_start = 0
h_end = image_size_h
elif index == 1: # Translate right 20 percent
offset = np.array([0.0, -0.3], dtype=np.float32)
size = np.array([image_size_w, ceil(0.7 * image_size_h)], dtype=np.int32)
w_start = int(floor((1 - 0.7) * image_size_w))
w_end = image_size_w
h_start = 0
h_end = image_size_h
elif index == 2: # Translate top 20 percent
offset = np.array([0.3, 0.0], dtype=np.float32)
size = np.array([ceil(0.7 * image_size_w), image_size_h], dtype=np.int32)
w_start = 0
w_end = image_size_w
h_start = 0
h_end = int(ceil(0.7 * image_size_h))
else: # Translate bottom 20 percent
offset = np.array([-0.3, 0.0], dtype=np.float32)
size = np.array([ceil(0.7 * image_size_w), image_size_h], dtype=np.int32)
w_start = 0
w_end = image_size_w
h_start = int(floor((1 - 0.7) * image_size_h))
h_end = image_size_h
return offset, size, w_start, w_end, h_start, h_end
def translate_images(X_imgs,image_size_w,image_size_h):
offsets = np.zeros((len(X_imgs), 2), dtype=np.float32)
n_translations = 4
X_translated_arr = []
tf.reset_default_graph()
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for i in range(n_translations):
X_translated = np.zeros((len(X_imgs), image_size_w,image_size_h, 3),
dtype=np.float32)
X_translated.fill(1.0) # Filling background color
base_offset, size, w_start, w_end, h_start, h_end = get_translate_parameters(i,image_size_w,image_size_h)
offsets[:, :] = base_offset
glimpses = tf.image.extract_glimpse(X_imgs, size, offsets)
glimpses = sess.run(glimpses)
X_translated[:, h_start: h_start + size[0], \
w_start: w_start + size[1], :] = glimpses
X_translated_arr.extend(X_translated)
X_translated_arr = np.array(X_translated_arr, dtype=np.float32)
return X_translated_arr
image = cv2.imread('plane.jpeg')
b,g,r = cv2.split(image)
image = cv2.merge([r,g,b])
image = cv2.resize(image, (200,200),0,0, cv2.INTER_LINEAR) #图像缩放
image = np.multiply(image, 1.0 / 255.0)
list_image = []
list_image.append(image)
list_image = np.array(list_image)
#scaled_imgs = central_scale_images(list_image, [1.30, 1.50, 1.80])
scaled_imgs = translate_images(list_image, 200,200)
#image = scaled_imgs[0]/255.
plt.subplot(231), plt.imshow(image), plt.title('original', fontsize='medium')
plt.subplot(232), plt.imshow(scaled_imgs[0]), plt.title('Translate left 20 percent', fontsize='medium')
plt.subplot(233), plt.imshow(scaled_imgs[1]), plt.title('Translate right 20 percent', fontsize='medium')
plt.subplot(234), plt.imshow(scaled_imgs[2]), plt.title('Translate top 20 percent', fontsize='medium')
plt.subplot(235), plt.imshow(scaled_imgs[3]), plt.title('Translate bottom 20 percent', fontsize='medium')
plt.show() 结果展示：
3.图片缩放
import numpy as np
import tensorflow as tf
import cv2
import matplotlib.pyplot as plt
#
# '''图片缩放'''
#
# # #image = tf.gfile.FastGFile('f-35-37.jpg','rb').read()
image = cv2.imread('plane.jpeg')
b,g,r = cv2.split(image)
image = cv2.merge([r,g,b])
image = cv2.resize(image, (200,200),0,0, cv2.INTER_LINEAR) #图像缩放
image = np.multiply(image, 1.0 / 255.0)
list_image = []
list_image.append(image)
def central_scale_images(X_imgs, scales,image_w=200,image_h=200):
# Various settings needed for Tensorflow operation
boxes = np.zeros((len(scales), 4), dtype=np.float32)
for index, scale in enumerate(scales):
x1 = y1 = 0.5 - 0.5 * scale # To scale centrally
x2 = y2 = 0.5 + 0.5 * scale
boxes[index] = np.array([y1, x1, y2, x2], dtype=np.float32)
box_ind = np.zeros((len(scales)), dtype=np.int32)
crop_size = np.array([image_w, image_h], dtype=np.int32)
X_scale_data = []
#tf.reset_default_graph()
X = tf.placeholder(tf.float32, shape=(1, image_w, image_h, 3))
# Define Tensorflow operation for all scales but only one base image at a time
tf_img = tf.image.crop_and_resize(X, boxes, box_ind, crop_size)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for img_data in X_imgs:
# img_data = np.multiply(image, 1.0 / 255.0) #归一化 np.multiply 乘法
batch_img = np.expand_dims(img_data, axis=0)
scaled_imgs = sess.run(tf_img, feed_dict={X: batch_img})
X_scale_data.extend(scaled_imgs)
#X_scale_data = np.array(X_scale_data, dtype=np.float32)
return X_scale_data
scaled_imgs = central_scale_images(list_image, [1.30, 1.50, 1.80])
#scaled_imgs = central_scale_images(list_image, [0.9, 0.7, 0.6])
#image = scaled_imgs[0]/255.
image2 = scaled_imgs[0]
plt.subplot(221), plt.imshow(image), plt.title('original')
plt.subplot(222), plt.imshow(image2), plt.title('1.30')
plt.subplot(223), plt.imshow(scaled_imgs[1]), plt.title('1.50')
plt.subplot(224), plt.imshow(scaled_imgs[2]), plt.title('1.80')
plt.show() 结果展示：
京公网安备 11010502036488号