深度学习之数据集分配

最新推荐文章于 2024-06-13 09:30:00 发布

Xnion

最新推荐文章于 2024-06-13 09:30:00 发布

阅读量341

点赞数 1

文章标签：数据集训练集验证集文件复制随机分布

本文链接：https://blog.csdn.net/Xnion/article/details/126010173

版权

import numpy as np
import os
import shutil

##### Image
path = "D:/Data/"
file_path0 = path + "image"                      #更换数据集 image 的文件名
path_list0 = os.listdir(file_path0)                 #会历遍文件夹内的文件并返回一个列表
path_name0 = []

for i in path_list0:
    path_name0.append(file_path0 + "/" + i)
    
# 排序一下  image
path_name0.sort()
train_path0 = []
trains_idx0 = []
val_path0 = []
vals_idx0 = []

##### label
file_path1 = path  + "label"                      #更换你数据集 label 的文件名
path_list1 = os.listdir(file_path1)                 #会历遍文件夹内的文件并返回一个列表
path_name1 = []
 
for j in path_list1:
    path_name1.append(file_path1 + "/" + j)
    
# 排序一下 label
path_name1.sort()
train_path1 = []
trains_idx1 = []
val_path1 = []
vals_idx1 = []


i = 0
for i in range(1):             #图片的类别数为1    
    idx = np.arange(0, 9960)
    np.random.shuffle(idx)
                                    # image
    train_idx0 = idx[0:7968]       #训练集每一类选取随机排序后  80%张作为训练集     
    val_idx0 = idx[7968:9960]         #验证集每一类选取随机排序后  20%作为验证集合
    trains_idx0.extend(train_idx0)
    vals_idx0.extend(val_idx0)
                                    # label
    train_idx1 = idx[0:7968]       #训练集每一类选取随机排序后  80%作为训练集     
    val_idx1 = idx[7968:9960]         #验证集每一类选取随机排序后  20%作为验证集合
    trains_idx1.extend(train_idx1)
    vals_idx1.extend(val_idx1)

# 如果没有存储分配数据后的文件夹，则创建文件夹
dirs = path + "training/train"
if not os.path.exists(dirs):
    os.mkdir(path + "training/train")
    os.mkdir(path + "training/val")
    os.mkdir(path + "training/train_label")
    os.mkdir(path + "training/val_label")

## 原图
path_name0 = np.array(path_name0)
train_path0 = path_name0[trains_idx0]
val_path0 = path_name0[vals_idx0]

for file_name0 in train_path0:
    shutil.copy(file_name0, path + "training/train")       #存放训练集文件的文件夹
for file_name0 in val_path0:
    shutil.copy(file_name0, path + "training/val")         #存放验证集文件的文件夹


## 标签
path_name1 = np.array(path_name1)
train_path1 = path_name1[trains_idx1]
val_path1 = path_name1[vals_idx1]

for file_name1 in train_path1:
    shutil.copy(file_name1, path + "training/train_label")  #存放训练集标签文件的文件夹
for file_name1 in val_path1:
    shutil.copy(file_name1, path + "training/val_label")    #存放验证集标签文件的文件夹
    
print("数据集分配完成！")