使用Tensorflow制作球鞋识别模型（二）——图片输入处理

最新推荐文章于 2021-06-27 19:42:15 发布

WellTung_666

最新推荐文章于 2021-06-27 19:42:15 发布

阅读量1.4k

点赞数 2

分类专栏： Tensorflow 文章标签： Tensorflow 数据处理图像识别

本文链接：https://blog.csdn.net/WellTung_666/article/details/88997986

版权

Tensorflow 专栏收录该内容

10 篇文章 3 订阅

订阅专栏

在上一篇博客使用Tensorflow制作球鞋识别模型（一）——数据预处理中已经介绍了TFRecord以及如何制作自己的图片数据集，接下来要对输入的图片进行处理，主要是实现对指定大小的生成图片进行sample与label分类制作获得神经网络输入的get_files文件，同时为了方便网络的训练，输入数据进行batch处理。

生成图片路径和标签的List

首先导入所需要的包

import os
import math
import numpy as np
import tensorflow as tf

图片路径和创建列表

train_dir = 'D:/PyCharm/PycharmProjects/AJ_Recognition/data_prepare/pic/inputdata'

AJ1 = []
label_AJ1 = []
AJ4 = []
label_AJ4 = []
AJ11 = []
label_AJ11 = []
AJ12 = []
label_AJ12 = []

获取路径下所有的图片路径名，存放到对应的列表中，同时贴上标签，存放到label列表中。
append() 方法用于在列表末尾添加新的对象。

def get_files(file_dir, ratio):
    for file in os.listdir(file_dir + '/AJ1'):
        AJ1.append(file_dir + '/AJ1' + '/' + file)
        label_AJ1.append(0)
    for file in os.listdir(file_dir + '/AJ4'):
        AJ4.append(file_dir + '/AJ4' + '/' + file)
        label_AJ4.append(1)
    for file in os.listdir(file_dir + '/AJ11'):
        AJ11.append(file_dir + '/AJ11' + '/' + file)
        label_AJ11.append(2)
    for file in os.listdir(file_dir + '/AJ12'):
        AJ12.append(file_dir + '/AJ12' + '/' + file)
        label_AJ12.append(3)

np.hstack()：按水平方向（列顺序）堆叠数组构成一个新的数组
np.array()：构造函数
temp.transpose()：转置
np.random.shuffle()：打乱顺序
math.ceil(x)：返回大于等于参数x的最小整数

    # 对生成的图片路径和标签List做打乱处理把各类合起来组成一个list
    image_list = np.hstack((AJ1, AJ4, AJ11, AJ12))
    label_list = np.hstack((label_AJ1, label_AJ4, label_AJ11, label_AJ12))

    # 利用shuffle打乱顺序
    temp = np.array([image_list, label_list])
    temp = temp.transpose()
    np.random.shuffle(temp)

    # 将所有的img和lab转换成list
    all_image_list = list(temp[:, 0])
    all_label_list = list(temp[:, 1])

    # 将所得List分为两部分，一部分用来训练，一部分用来测试，ratio是测试集的比例
    n_sample = len(all_label_list)
    
    # 测试样本数
    n_val = int(math.ceil(n_sample * ratio))  
    
    # 训练样本数
    n_train = n_sample - n_val 

    tra_images = all_image_list[0:n_train]
    tra_labels = all_label_list[0:n_train]
    tra_labels = [int(float(i)) for i in tra_labels]
    val_images = all_image_list[n_train:-1]
    val_labels = all_label_list[n_train:-1]
    val_labels = [int(float(i)) for i in val_labels]

    return tra_images, tra_labels, val_images, val_labels

生成Batach

def get_batch(image, label, image_W, image_H, batch_size, capacity):

    # 转换类型
    image = tf.cast(image, tf.string)
    label = tf.cast(label, tf.int32)

    # 输入队列
    input_queue = tf.train.slice_input_producer([image, label])
    label = input_queue[1]
    image_contents = tf.read_file(input_queue[0])  # read img from a queue

    #将图像解码，不同类型的图像不能混在一起
    image = tf.image.decode_jpeg(image_contents, channels=3)

    # 数据预处理，对图像进行旋转、缩放、裁剪、归一化等操作
    image = tf.image.resize_image_with_crop_or_pad(image, image_W, image_H)
    image = tf.image.per_image_standardization(image)

    #生成batch
    image_batch, label_batch = tf.train.batch([image, label],batch_size=batch_size,capacity=capacity)
    
    # 重新排列label，行数为[batch_size]
    label_batch = tf.reshape(label_batch, [batch_size])
    image_batch = tf.cast(image_batch, tf.float32)
    return image_batch, label_batch

全部代码

import os
import math
import numpy as np
import tensorflow as tf

train_dir = 'D:/PyCharm/PycharmProjects/AJ_Recognition/data_prepare/pic/inputdata'

AJ1 = []
label_AJ1 = []
AJ4 = []
label_AJ4 = []
AJ11 = []
label_AJ11 = []
AJ12 = []
label_AJ12 = []

def get_files(file_dir, ratio):
    for file in os.listdir(file_dir + '/AJ1'):
        AJ1.append(file_dir + '/AJ1' + '/' + file)
        label_AJ1.append(0)
    for file in os.listdir(file_dir + '/AJ4'):
        AJ4.append(file_dir + '/AJ4' + '/' + file)
        label_AJ4.append(1)
    for file in os.listdir(file_dir + '/AJ11'):
        AJ11.append(file_dir + '/AJ11' + '/' + file)
        label_AJ11.append(2)
    for file in os.listdir(file_dir + '/AJ12'):
        AJ12.append(file_dir + '/AJ12' + '/' + file)
        label_AJ12.append(3)

    image_list = np.hstack((AJ1, AJ4, AJ11, AJ12))
    label_list = np.hstack((label_AJ1, label_AJ4, label_AJ11, label_AJ12))

    temp = np.array([image_list, label_list])
    temp = temp.transpose()
    np.random.shuffle(temp)

    all_image_list = list(temp[:, 0])
    all_label_list = list(temp[:, 1])
    
    n_sample = len(all_label_list)
    n_val = int(math.ceil(n_sample * ratio)) 
    n_train = n_sample - n_val  

    tra_images = all_image_list[0:n_train]
    tra_labels = all_label_list[0:n_train]
    tra_labels = [int(float(i)) for i in tra_labels]
    val_images = all_image_list[n_train:-1]
    val_labels = all_label_list[n_train:-1]
    val_labels = [int(float(i)) for i in val_labels]

    return tra_images, tra_labels, val_images, val_labels

def get_batch(image, label, image_W, image_H, batch_size, capacity):
    image = tf.cast(image, tf.string)
    label = tf.cast(label, tf.int32)
    
    input_queue = tf.train.slice_input_producer([image, label])
    
    label = input_queue[1]
    
    image_contents = tf.read_file(input_queue[0])  
    
    image = tf.image.decode_jpeg(image_contents, channels=3)
    image = tf.image.resize_image_with_crop_or_pad(image, image_W, image_H)
    image = tf.image.per_image_standardization(image)
    
    image_batch, label_batch = tf.train.batch([image, label],
                                              batch_size=batch_size,
                                              num_threads=32,
                                              capacity=capacity)    
                                                                                     
    label_batch = tf.reshape(label_batch, [batch_size])
    image_batch = tf.cast(image_batch, tf.float32)
    
    return image_batch, label_batch