import numpy as np
import torch
import os
import random
from PIL import Image
from torch.utils.data import Dataset
from torch.utils.data import DataLoader
import torchvision.transforms as transforms
BASE_PATH = os.path.abspath(__file__)
# print(BASE_PATH)
base_path = os.path.abspath(os.path.join(BASE_PATH, '..', 'TestDir'))
# print(base_path)
data_dir = os.path.abspath(os.path.join(BASE_PATH, '..', 'RMB_data'))
random.seed(1)
# print(data_dir)
test_label = {"1": 0, "100": 1}
data_info = list()
for path, dirs, files in os.walk(base_path):
for sub_dir in dirs:
# print(sub_dir)
sub_dirlist = os.listdir(os.path.join(base_path, sub_dir))
pynames = list(filter(lambda y: y.endswith('.jpg'), sub_dirlist))
# print(pynames)
# print(test_label[sub_dir])
for pyname in pynames:
datainfo_dir = os.path.join(base_path, sub_dir, pyname)
t_label=test_label[sub_dir]
t_label = int(t_label)
data_info.append((datainfo_dir, t_label))
# print(data_info)
new_data_info = list()
for data_info_e in data_info:
x_dir, x_label = data_info_e
x_img = Image.open(x_dir).convert('RGB')
ok_transform = transforms.Compose([
transforms.Resize((32, 32)),
transforms.RandomCrop(32, padding=4),
transforms.ToTensor(),
])
x_img = ok_transform(x_img)
new_data_info.append((x_img,x_label))
# print(len(new_data_info[0][0]))
print(len(new_data_info))
newdataLoader = DataLoader(new_data_info,batch_size=14, shuffle=True)
for ids, data in enumerate(newdataLoader):
print(ids)
pytorch(08)数据模型的读取(2)
最新推荐文章于 2024-07-10 23:51:30 发布