将图片按txt格式进行存储，对txt文件批量读取

为什么将图片存储为txt格式

在将图片数据输入卷积神经网络时，如果通过np.load进行输入，常常会出现内存爆炸的问题，当然啊，土豪除外。所以讲图片按照txt文件读取，是一种非常好的读取方式，当然啊，如今ternsorflow出的API也相当不错。技多不压身嘛！

将图片存储为txt文件代码

'''
本代码以UC数据集为例子，UC数据集共包含21类，每一类图片包含100张图片
其中class_label_dict为UC数据集label，
'''
import numpy as np
import os


class_label_dict = {'agricultural': 0,
                   "airplane": 1,
                   "baseballdiamond": 2,
                   "beach": 3,
                   "buildings": 4,
                   "chaparral": 5,
                   "denseresidential": 6,
                   "forest": 7,
                   "freeway": 8,
                   "golfcourse": 9,
                   "harbor": 10,
                   "intersection": 11,
                   "mediumresidential": 12,
                   "mobilehomepark": 13,
                   "overpass": 14,
                   "parkinglot": 15,
                   "river": 16,
                   "runway": 17,
                   "sparseresidential": 18,
                   "storagetanks": 19,
                   "tenniscourt": 20
                   }
file_path = "G:/UC"                                  #此处为文件存放位置
path_list = os.listdir(file_path)                 #会历遍文件夹内的文件并返回一个列表
path_name=[]

for i in path_list:
   path_name.append(file_path+"/"+i+" "+str(class_label_dict[i[:-6]]))

# 排序一下
path_name.sort()


train_path = []
test_path = []
trains_idx = []
tests_idx = []
for i in range(21):
   start = i * 100
   end = (i + 1) * 100
   idx = np.arange(start, end)
   np.random.shuffle(idx)
   train_idx = idx[0:80]
   test_idx = idx[80:]
   trains_idx.extend(train_idx)
   tests_idx.extend(test_idx)
path_name = np.array(path_name)
train_path = path_name[trains_idx]
test_path = path_name[tests_idx]
for file_name in path_name:
   # "a"表示以不覆盖的形式写入到文件中,当前文件夹如果没有"save.txt"会自动创建
   with open("data.txt", "a") as f:
       f.write(file_name + "\n")
   f.close()

for file_name in train_path:
   # "a"表示以不覆盖的形式写入到文件中,当前文件夹如果没有"save.txt"会自动创建
   with open("train.txt", "a") as f:
       f.write(file_name + "\n")
   f.close()
for file_name in test_path:
   # "a"表示以不覆盖的形式写入到文件中,当前文件夹如果没有"save.txt"会自动创建
   with open("test.txt", "a") as f:
       f.write(file_name + "\n")
   f.close()

输出txt文件的格式为 文件位置+文件名+label形式 ，如下图所示：

将txt文件进行批量读取

def read_txt():
   train_filename = []
   train_filelabel = []
   with open('/home/admin324/PycharmProjects/qiushuo/A_GS/data/train.txt', 'r') as f:
       x = f.readlines()
       for name in x:
           train_filename.append(name.strip().split()[0])
           train_filelabel.append(int(name.strip().split()[1]))
   train_filename = np.array(train_filename)
   train_filelabel = np.array(train_filelabel)

   test_filename = []
   test_filelabel = []
   with open('/home/admin324/PycharmProjects/qiushuo/A_GS/data/test.txt', 'r') as f:
       x = f.readlines()
       for name in x:
           test_filename.append(name.strip().split()[0])
           test_filelabel.append(int(name.strip().split()[1]))
   test_filename = np.array(test_filename)
   test_filelabel = np.array(test_filelabel)
   return train_filename, train_filelabel, test_filename, test_filelabel


def read_data(path):
   data = []
   for i in range(path.shape[0]):
       temp = cv2.imread(path[i])
       temp = cv2.cvtColor(temp, cv2.COLOR_BGR2RGB)
       temp = cv2.resize(temp, dsize=(224, 224))
       data.append(temp / 255.)
   data = np.array(data)
   return data