pytorch引用halcon写数据集

****加粗样式虽然啰嗦一点，但好歹halcon自己熟悉，不会忘记，用os 和 pil会导致脑子记得东西太多

import halcon as ha
import torch
from torch.utils.data import Dataset

path0 = r'E:\BaiduNetdiskDownload\cell'
class MyDataset(Dataset):
    def __init__(self, path0):
        self.path = path0
        self.image_files = ha.list_files(path0 + '\masks','files')
    def __len__(self):
        number= ha.tuple_length(self.image_files)
        return number
    def __getitem__(self, index):
        path1=self.image_files[index]
        print('path1',path1)
        image_marks = ha.read_image(path1)
        #### 这里要注意的是分割符  “\\”
        tuple_string = ha.tuple_split(path1,"\\")
        leng = ha.tuple_length(tuple_string)
        name_string =ha.tuple_select(tuple_string,leng-1)
        name_s = ha.tuple_split(name_string,'.')
        name = name_s[0]

        test_image_files = ha.list_files(self.path + '\imgs','files')
        path = ha.tuple_regexp_select(test_image_files, name+'.')
        ###如果数据集是0,1.。。。。200这样的话 0,10,100,110是比较麻烦的，建议命名用 n+数字  也就是英文字母加数字的形式更方便一些
        if int(name) <10:
            path=path[0]
        print(path)
        image_text =ha.read_image(path)

        imagetext_np = ha.himage_as_numpy_array(image_text)
        image_marks_np = ha.himage_as_numpy_array(image_marks)
        return  imagetext_np,image_marks_np,

dataset = MyDataset(path0)
print(dataset.__len__())
text,marks =dataset.__getitem__(0)
print('text',text)
print('marks',marks)