Python torchvision.datasets.ImageNet() Examples

The following are 8 code examples for showing how to use torchvision.datasets.ImageNet(). These examples are extracted from open source projects. You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example.

You may check out the related API usage on the sidebar.

You may also want to check out all available functions/classes of the module torchvision.datasets , or try the search function .

Example 1
Project: Deep_Openset_Recognition_through_Uncertainty   Author: MrtnMndt   File:    License: MIT License 6 votes vote down vote up
def get_dataset(self):
        Uses torchvision.datasets.ImageNet to load dataset.
        Downloads dataset if doesn't exist already.
    trainset, valset

        trainset = datasets.ImageNet('datasets/ImageNet/train/', split='train', transform=self.train_transforms,
                                     target_transform=None, download=True)
        valset = datasets.ImageNet('datasets/ImageNet/val/', split='val', transform=self.val_transforms,
                                   target_transform=None, download=True)

        return trainset, valset 
Example 2
Project: MobileNetV3-pytorch   Author: Randl   File:    License: MIT License 5 votes vote down vote up
def get_loaders(dataroot, val_batch_size, train_batch_size, input_size, workers, num_nodes, local_rank):
    # TODO: pin-memory currently broken for distributed
    pin_memory = False
    # TODO: datasets.ImageNet
    val_data = datasets.ImageFolder(root=os.path.join(dataroot, 'val'), transform=get_transform(False, input_size))
    val_sampler = DistributedSampler(val_data, num_nodes, local_rank)
    val_loader =, batch_size=val_batch_size, sampler=val_sampler,
                                             num_workers=workers, pin_memory=pin_memory)

    train_data = datasets.ImageFolder(root=os.path.join(dataroot, 'train'),
    train_sampler = DistributedSampler(train_data, num_nodes, local_rank)
    train_loader =, batch_size=train_batch_size, sampler=train_sampler,
                                               num_workers=workers, pin_memory=pin_memory)
    return train_loader, val_loader 
Example 3
Project: sotabench-eval   Author: paperswithcode   File:    License: Apache License 2.0 5 votes vote down vote up
def load_targets(self):
        Downloads ImageNet labels and IDs and puts into self.root, then loads to self.targets

        :return: void - update self.targets with the ImageNet validation data labels, and downloads if
        the pickled validation data is not in the root location

        with open(os.path.join(self.root, 'imagenet_val_targets.pkl'), 'rb') as handle:
            self.targets = pickle.load(handle) 
Example 4
Project: sotabench-eval   Author: paperswithcode   File:    License: Apache License 2.0 5 votes vote down vote up
def save(self):
        Calculate results and then put into a BenchmarkResult object

        On the server, this will produce a JSON file serialisation in sotabench_results.json and results
        will be recorded on the platform.

        :return: BenchmarkResult object with results and metadata
        # recalculate to ensure no mistakes made during batch-by-batch metric calculation

        # If this is the first time the model is run, then we record evaluation time information

        if not self.cached_results:
            exec_speed = (time.time() - self.init_time)
            self.speed_mem_metrics['Tasks / Evaluation Time'] = len(self.outputs) / exec_speed
            self.speed_mem_metrics['Tasks'] = len(self.outputs)
            self.speed_mem_metrics['Evaluation Time'] = exec_speed
            self.speed_mem_metrics['Tasks / Evaluation Time'] = None
            self.speed_mem_metrics['Tasks'] = None
            self.speed_mem_metrics['Evaluation Time'] = None

        return BenchmarkResult(
Example 5
Project: homura   Author: moskomule   File:    License: Apache License 2.0 5 votes vote down vote up
def __init__(self,
        assert not download, "Download dataset by yourself!"
        super(ImageNet, self).__init__(root, split="train" if train else "val", transform=transform) 
Example 6
Project: sotabench-eval   Author: paperswithcode   File:    License: Apache License 2.0 4 votes vote down vote up
def get_results(self):
        Gets the results for the evaluator. This method only runs if predictions for all 5,000 ImageNet validation
        images are available. Otherwise raises an error and informs you of the missing or unmatched IDs.

        :return: dict with Top 1 and Top 5 Accuracy

        if self.cached_results:
            return self.results

        if set(self.targets.keys()) != set(self.outputs.keys()):
            missing_ids = set(self.targets.keys()) - set(self.outputs.keys())
            unmatched_ids = set(self.outputs.keys()) - set(self.targets.keys())

            if len(unmatched_ids) > 0:
                raise ValueError('''There are {mis_no} missing and {un_no} unmatched image IDs\n\n'''
                                     '''Missing IDs are {missing}\n\n'''
                                     '''Unmatched IDs are {unmatched}'''.format(mis_no=len(missing_ids),
                raise ValueError('''There are {mis_no} missing image IDs\n\n'''
                                     '''Missing IDs are {missing}'''.format(mis_no=len(missing_ids),

        # Do the calculation only if we have all the results...
        self.top1 = AverageMeter()
        self.top5 = AverageMeter()

        for i, dict_key in enumerate(tqdm.tqdm(self.targets.keys())):
            output = self.outputs[dict_key]
            target = self.targets[dict_key]
            prec1 = top_k_accuracy_score(y_true=target, y_pred=np.array([output]), k=1)
            prec5 = top_k_accuracy_score(y_true=target, y_pred=np.array([output]), k=5)
            self.top1.update(prec1, 1)
            self.top5.update(prec5, 1)

        self.results = {'Top 1 Accuracy': self.top1.avg, 'Top 5 Accuracy': self.top5.avg}
        self.speed_mem_metrics['Max Memory Allocated (Total)'] = get_max_memory_allocated()

        return self.results 
Example 7
Project: scratchai   Author: iArunava   File:    License: MIT License 4 votes vote down vote up
def pre_benchmark_atk(**kwargs):
  Helper function that sets all the defaults while performing checks
  for all the options passed before benchmarking attacks.

  # Set the Default options if nothing explicit provided
  def_dict = {
    'bs'       : 4,
    'trf'      : get_trf('rz256_cc224_tt_normimgnet'),
    'dset'     : 'NA',
    'root'     : './',
    'topk'     : (1, 5),
    'dfunc'    : datasets.ImageFolder,
    'download' : True,

  for key, val in def_dict.items(): 
    if key not in kwargs: kwargs[key] = val
  if kwargs['dset'] == 'NA':
    if 'loader' not in kwargs:
      dset = kwargs['dfunc'](kwargs['root'], transform=kwargs['trf'])
      loader = DataLoader(dset, batch_size=kwargs['bs'], num_workers=2)
      loader = kwargs['loader']
  # Set dataset specific functions here
    if kwargs['dset'] == IMGNET12:
      dset = datasets.ImageNet(kwargs['root'], split='test',
                      download=kwargs['download'], transform=kwargs['trf'])
    elif kwargs['dset'] == MNIST:
      kwargs['trf'] = get_trf('tt_normmnist')
      kwargs['dfunc'] = datasets.MNIST
      dset = kwargs['dfunc'](kwargs['root'], train=False, 
                     download=kwargs['download'], transform=kwargs['trf'])
    else: raise

    loader = DataLoader(dset, shuffle=False, batch_size=kwargs['bs'])
  topk = kwargs['topk']
  for key, val in kwargs.items():
    print ('[INFO] Setting {} to {}.'.format(key, kwargs[key]))

  # Deleting keys that is used just for benchmark_atk() function is 
  # important as the same kwargs dict is passed to initialize the attack
  # So, otherwise the attack will throw an exception
  for key in def_dict:
    del kwargs[key]
  if 'loader' in kwargs: del kwargs['loader']

  return loader, topk, kwargs 
Example 8
Project: ignite   Author: pytorch   File:    License: BSD 3-Clause "New" or "Revised" License 4 votes vote down vote up
def get_train_val_loaders(
    root_path: str,
    train_transforms: Callable,
    val_transforms: Callable,
    batch_size: int = 16,
    num_workers: int = 8,
    val_batch_size: Optional[int] = None,
    limit_train_num_samples: Optional[int] = None,
    limit_val_num_samples: Optional[int] = None,
) -> Tuple[DataLoader, DataLoader, DataLoader]:

    train_ds = ImageNet(
        root_path, split="train", transform=lambda sample: train_transforms(image=sample)["image"], loader=opencv_loader
    val_ds = ImageNet(
        root_path, split="val", transform=lambda sample: val_transforms(image=sample)["image"], loader=opencv_loader

    if limit_train_num_samples is not None:
        train_indices = np.random.permutation(len(train_ds))[:limit_train_num_samples]
        train_ds = Subset(train_ds, train_indices)

    if limit_val_num_samples is not None:
        val_indices = np.random.permutation(len(val_ds))[:limit_val_num_samples]
        val_ds = Subset(val_ds, val_indices)

    # random samples for evaluation on training dataset
    if len(val_ds) < len(train_ds):
        train_eval_indices = np.random.permutation(len(train_ds))[: len(val_ds)]
        train_eval_ds = Subset(train_ds, train_eval_indices)
        train_eval_ds = train_ds

    train_loader = idist.auto_dataloader(
        train_ds, shuffle=True, batch_size=batch_size, num_workers=num_workers, drop_last=True,

    val_batch_size = batch_size * 4 if val_batch_size is None else val_batch_size
    val_loader = idist.auto_dataloader(
        val_ds, shuffle=False, batch_size=val_batch_size, num_workers=num_workers, drop_last=False,

    train_eval_loader = idist.auto_dataloader(
        train_eval_ds, shuffle=False, batch_size=val_batch_size, num_workers=num_workers, drop_last=False,

    return train_loader, val_loader, train_eval_loader