Torch DeepHash Dataset

A simple PyTorch implementation of common datasets for Deep Hashing. For now, only NUS-WIDE and COCO datasets are supported.

Installation

Install via pip

pip install git+https://github.com/nghia-ndx/torch-deephash-dataset

Usages

• Each dataset supports 3 types of splits:
  • train: Train set
  • test: Test set
  • db: Full dataset (exclude train and test)
• On first __init__, the dataset will be downloaded to the specified root location and will be reused on future runs.

from torch_deephash_dataset.coco import COCODataset
from torch_deephash_dataset.nus_wide import NUSWIDEDataset

coco_dataset = COCODataset('datasets/coco', split='train')
nus_wide_dataset = NUSWIDEDataset('datasets/nus_wide', split='test')

• Subsequently, if you want to re-download, use force_download=True.

  Note: using force_download=True will DELETE everything in the specified root directory before re-downloading.

from torch_deephash_dataset.nus_wide import NUSWIDEDataset

nus_wide_dataset = NUSWIDEDataset(
    'datasets/nus_wide', 
    split='test', 
    force_download=True
)

• The dataset could be used with PyTorch's DataLoader as usual:

from torch.utils.data import DataLoader

data_loader = DataLoader(
    dataset=nus_wide_dataset, 
    batch_size=64, 
    num_workers=4
)

• Data transform and label transform can be appied via transform and target_tranform arguments:

from torch_deephash_dataset.nus_wide import NUSWIDEDataset
from torchvision import transforms

nus_wide_dataset = NUSWIDEDataset(
    'datasets/nus_wide', 
    split='test', 
    transform=transforms.ToTensor()
)