【图像分类】将目标文件数据划分训练集、验证集和测试集三个文件夹，并加载数据

 一、训练集、验证集和测试集三个文件夹的划分：

代码：

import os
import random
from shutil import copy2


def data_set_split(src_data_folder, target_data_folder, slice_data = [0.4,0.3,0.3]):
    '''
    读取源数据文件夹，生成划分好的文件夹，分为trian、val、test三个文件夹进行
    :param src_data_folder: r"D:\Desktop\segmentation_2021\data"
    :param target_data_folder: 目标文件夹 r"D:\Desktop\segmentation_2021\a"
    :param slice_data: 划分数据比例比例  训练 验证 测试所占百分比
    :return:
    '''
    print("开始数据集划分")
    class_names = os.listdir(src_data_folder)
    # 在目标目录下创建文件夹
    split_names = ['train', 'val', 'test']
    for split_name in split_names:
        split_path = os.path.join(target_data_folder, split_name)
        if os.path.isdir(split_path):
            pass
        else:
            os.mkdir(split_path)
        # 然后在split_path的目录下创建类别文件夹
        for class_name in class_names:
            class_split_path = os.path.join(split_path, class_name)
            if os.path.isdir(class_split_path):
                pass
            else:
                os.mkdir(class_split_path)

    # 按照比例划分数据集，并进行数据图片的复制
    # 首先进行分类遍历
    for class_name in class_names:
        current_class_data_path = os.path.join(src_data_folder, class_name)
        current_all_data = os.listdir(current_class_data_path)
        current_data_length = len(current_all_data)
        current_data_index_list = list(range(current_data_length))
        random.shuffle(current_data_index_list)

        train_folder = os.path.join(os.path.join(target_data_folder, 'train'), class_name)
        val_folder = os.path.join(os.path.join(target_data_folder, 'val'), class_name)
        test_folder = os.path.join(os.path.join(target_data_folder, 'test'), class_name)
        train_stop_flag = current_data_length * slice_data[0]
        val_stop_flag = current_data_length * (slice_data[0] + slice_data[1])
        current_idx = 0
        train_num = 0
        val_num = 0
        test_num = 0
        for i in current_data_index_list:
            src_img_path = os.path.join(current_class_data_path, current_all_data[i])
            if current_idx <= train_stop_flag:
                copy2(src_img_path, train_folder)
                # print("{}复制到了{}".format(src_img_path, train_folder))
                train_num = train_num + 1
            elif (current_idx > train_stop_flag) and (current_idx <= val_stop_flag):
                copy2(src_img_path, val_folder)
                # print("{}复制到了{}".format(src_img_path, val_folder))
                val_num = val_num + 1
            else:
                copy2(src_img_path, test_folder)
                # print("{}复制到了{}".format(src_img_path, test_folder))
                test_num = test_num + 1

            current_idx = current_idx + 1

        print("*********************************{}*************************************".format(class_name))
        print(
            "{}类按照{}：{}：{}的比例划分完成，一共{}张图片".format(class_name, slice_data[0], slice_data[1], slice_data[2], current_data_length))
        print("训练集{}：{}张".format(train_folder, train_num))
        print("验证集{}：{}张".format(val_folder, val_num))
        print("测试集{}：{}张".format(test_folder, test_num))


if __name__ == '__main__':
    src_data_folder = r"D:\Desktop\segmentation_2021\data"
    target_data_folder = r"D:\Desktop\segmentation_2021\a"
    data_set_split(src_data_folder, target_data_folder, slice_data=[0.6,0.2,0.2])

参考：https://zhuanlan.zhihu.com/p/199238910

 二、数据加载

         划分之后的数据加载。

代码：

import torch.utils.data as Data
import numpy as np
import torchvision
import matplotlib.pyplot as plt

data_transform = torchvision.transforms.Compose([
    torchvision.transforms.Resize((128,128)), # 将图片的尺寸 Resize 到128*128
    torchvision.transforms.ToTensor(), # 转为张量并归一化到[0,1]（是将数据除以255），且会把H*W*C会变成C *H *W
    torchvision.transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225],)]) # 数据归一化处理
path = r"D:\Desktop\train"  # train数据的上一层目录（即数据的上一层目录）
train_data = torchvision.datasets.ImageFolder(root=path,transform=data_transform)

train_data_loader = Data.DataLoader(
    dataset=train_data,
    batch_size=32,
    shuffle=True,
)