kaggle:code 猫狗识别 图像识别TensorFlow图像预处理

#coding:utf8

import matplotlib.pyplot as plt
import numpy as np
import os
import tensorflow as tf
#import cv2
TRAIN_DIR = './wtrain/'
TEST_DIR = './wtest/'


#提取出前2000的训练集样例
#提取出前1000的测试集样例
train_image_file_names = [TRAIN_DIR+i for i in os.listdir(TRAIN_DIR)][0:2000]
test_image_file_names = [TEST_DIR+i  for i in os.listdir(TEST_DIR)][0:1000]


#tensorflow计算图
def decode_image(image_file_names,resize_func=None):
    images = []

    #读取文件，将文件进行预处理成tensorflow可识别格式
    graph = tf.Graph()
    with graph.as_default():
        file_name = tf.placeholder(dtype=tf.string)
        file = tf.read_file(file_name)
        image = tf.image.decode_jpeg(file)
        if resize_func != None:
            image = resize_func(image)

    #启动会话层，讲转化后的图片结果添加到列表images中
    with tf.Session(graph=graph) as session:
        tf.global_variables_initializer().run()
        for i in range(len(image_file_names)):
            images.append(session.run(image,feed_dict={file_name:image_file_names[i]}))
            if (i+1)%1000 == 0:
                print ('Images proccessed:',i+1)

        session.close()
    return images


train_images = decode_image(train_image_file_names)
test_images = decode_image(test_image_file_names)
all_images = train_images + test_images

width = []
height = []
aspect_ratio = []
for image in all_images:
    h,w,d = np.shape(image)
    aspect_ratio.append(float(w)/float(h))
    width.append(w)
    height.append(h)

print ("Mean aspect ratio:",np.mean(aspect_ratio))
plt.plot(aspect_ratio)
plt.show()

print ('Mean width : ',np.mean(width))
print ('Mean height : ',np.mean(height))
plt.plot(width,height,'.r')
plt.show()

print("Images widther than 500 pixel: ", np.sum(np.array(width) > 500))
print("Images higher than 500 pixel: ", np.sum(np.array(height) > 500))

del train_images
del test_images
del all_images


#图像转化resize
WIDTH = 500
HEIGHT = 500
resize_func = lambda image: tf.image.resize_image_with_crop_or_pad(image,HEIGHT,WIDTH)

processed_train_images = decode_image(train_image_file_names,resize_func=resize_func)
processed_test_images = decode_image(test_image_file_names,resize_func=resize_func)

print (np.shape(processed_train_images))
print (np.shape(processed_test_images))

for i in range(10):
    plt.imshow(processed_train_images[i])
    plt.show()

import pickle

def create_batch(data,label,batch_size):
    i = 0
    while i*batch_size <= len(data):
        with open(label+'_'+str(i)+'.pickle','wb') as handle:
            content = data[(i*batch_size):((i+1)*batch_size)]
            pickle.dump(content,handle)
            print ('Saved',label,'part #' + str(i),'with',len(content),'entries.')
            i += 1

#创建one-hot数据类型
labels = [[1.0,0.0] if 'dog' in name else [0.0,1.0] for name in train_image_file_names]