如何在PyTorch中加载具有多个JSON注释的图像

Question

我想知道如何在PyTorch中将数据加载器用于我的自定义文件结构。我已经阅读了PyTorch文档，但是所有这些都带有单独的类文件夹。

我的文件夹结构由2个文件夹（称为训练和验证）组成，每个文件夹具有2个子文件夹（称为images和json_annotations）。 “ images”文件夹中的每个图像都有多个对象（例如汽车，自行车，人等），并且每个对象都带有注释，并具有单独的JSON文件。遵循标准的可可注释。我的意图是制作一个可以从视频进行实时分类的神经网络。

编辑1： 我已经按照FábioPerez的建议进行了编码。

class lDataSet(data.Dataset):
    def __init__(self, path_to_imgs, path_to_json):
        self.path_to_imgs = path_to_imgs
        self.path_to_json = path_to_json
        self.img_ids = os.listdir(path_to_imgs)

    def __getitem__(self, idx):
        img_id = self.img_ids[idx]
        img_id = os.path.splitext(img_id)[0]
        img = cv2.imread(os.path.join(self.path_to_imgs, img_id + ".jpg"))
        load_json = json.load(open(os.path.join(self.path_to_json, img_id + ".json")))
        #n = len(load_json)
        #bboxes = load_json['annotation'][n]['segmentation']
        return img, load_json

    def __len__(self):
        return len(self.image_ids)

当我尝试

l_data = lDataSet(path_to_imgs = '/home/training/images', path_to_json = '/home/training/json_annotations')

我用l_data [] [0]-图像获取l_data和使用json的l_data 。现在我很困惑。如何在PyTorch中通过微调示例availalbe使用它？在该示例中，数据集和数据加载器如下所示。
https://pytorch.org/tutorials/beginner/finetuning_torchvision_models_tutorial.html

# Create training and validation datasets
image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x), data_transforms[x]) for x in ['train', 'val']}
# Create training and validation dataloaders
dataloaders_dict = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=batch_size, shuffle=True, num_workers=4) for x in ['train', 'val']}

Answer 1

您应该可以使用data.Dataset来实现自己的数据集。您只需要实现__name__ = 'DEMO' class GradeHandler(object): EMNER = ["INFO100","INFO104","INFO110","INFO150","INFO125"] FAGKODER= [["Informasjonsvitenskap","INF"],["Kognitiv vitenskap","KVT"]] KARAKTERER = [["INFO100","C"],["INFO104","B"],["INFO110","E"]] def __init__(self): self.Emner = self.EMNER self.FagKoder = self.FAGKODER self.Karakterer = self.KARAKTERER self.__create_grade_dict() def remove_subject(self, subject_name): """ Remove a subject ot the classes class list variable. """ try: self.Emner = [i for i in self.EMNER if i != subject_name] self.__create_grade_dict() except ValueError: pass def add_subject(self, subject_name): """ Append a subject ot the classes class list variable. """ if not subject_name in Emner: self.Emner.append(subject_name) self.__create_grade_dict() def __create_grade_dict(self, grade_dict=None): """ Split grades matrix into separate parts; Create and set a dictionary of values. """ if grade_dict is None: self.grade_dict = dict() sub, grade = zip(*self.Karakterer) karakterer_dict = {k:v for k, v in list(zip(sub, grade))} for i in self.Emner: if i in karakterer_dict.keys(): self.grade_dict[i] = karakterer_dict[i] else: self.grade_dict[i] = '' def update_grade(self, subject_name, grade='A'): """ Update a grade in the grade dictionary. Will also add a subject if not alrady in the dictionary. """ try: self.grade_dict[subject_name] = grade except (KeyError, ValueError): pass def print_grades(self, subject_name=None): """ Print dictionary results. """ if subject_name is None: for k, v in self.grade_dict.items(): print('{} {}'.format(k, v)) else: if subject_name in self.grade_dict.keys(): print('{} {}'.format(subject_name, self.grade_dict[subject_name])) if __name__ == 'DEMO': ### Create an instance of the GradeHandler and print initial grades. gh = GradeHandler() gh.print_grades() ### Append a class gh.add_subject('GE0124') gh.print_grades() ### Add grade gh.update_grade('GE0124', 'B+') gh.print_grades() ### Update grades gh.update_grade('GE0124', 'A-') gh.print_grades() ### Remove subject (will also remove grade. gh.remove_subject('GE0124') gh.print_grades() 和__len__方法。

根据您的情况，您可以遍历图像文件夹中的所有图像（然后可以将图像ID存储在__getitem__中的列表中）。然后，您使用传递给Dataset的索引来获取相应的图像ID。使用此图像ID，您可以读取相应的JSON文件并返回所需的目标数据。

类似这样的东西：

__getitem__

在class YourDataLoader(data.Dataset): def __init__(self, path_to_imgs, path_to_json): self.path_to_imags = path_to_imgs self.path_to_json = path_to_json self.image_ids = iterate_through_images(path_to_images) def __getitem__(self, idx): img_id = self.image_ids[idx] img = load_image(os.path.join(self.path_to_images, img_id) bboxes = load_bboxes(os.path.join(self.path_to_json, img_id) return img, bboxes def __len__(self): return len(self.image_ids) 中，您将获得目录中图像的所有ID（例如文件名）。 在iterate_through_images中，您将读取JSON并获取所需的信息。

如果您要引用，我有一个JSON加载程序实现here。