dataset.py

import json
import urllib.request

import torchvision.transforms as T
from cods.classif.data import ClassificationDataset
from PIL import Image


class DatasetWrapper(ClassificationDataset):
    def __init__(self, dataset, transforms=None, **kwargs):
        self.dataset = dataset
        self.root = "./data"
        path = self.root
        self.image_ids = []

        if transforms is None:
            transforms = T.Compose(
                [
                    T.Resize(256),
                    T.CenterCrop(224),
                    T.ToTensor(),
                    T.Normalize(
                        mean=[0.485, 0.456, 0.406],
                        std=[0.229, 0.224, 0.225],
                    ),
                ],
            )
        self.transforms = transforms
        tmp = json.loads(
            urllib.request.urlopen(
                "https://s3.amazonaws.com/deep-learning-models/image-models/imagenet_class_index.json",
            ).read(),
        )
        wdnids = {int(k): v[0] for k, v in tmp.items()}
        self.wdnids = wdnids
        idx_to_cls = {int(k): v[1] for k, v in tmp.items()}
        # super().__init__(
        #     path=path,
        #     transforms=transforms,
        #     idx_to_cls=idx_to_cls,
        #     **kwargs,
        # )
        self.idx_to_cls = idx_to_cls

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

    def __getitem__(self, idx):
        item = self.dataset[idx]
        # print(item)
        img, label = item["image"], item["label"]

        if img.mode != "RGB":
            img = img.convert("RGB")
        # img = Image.open(path)
        if self.transforms:
            img = self.transforms(img)
        return idx, img, label