TinyImageNetDataset

A Tiny Imagenet Dataset for PyTorch, the usage is similar to torchvision.datasets, this means each sample in the Dataset is a PIL.Image.Image.

This script will download the dataset from the internet to your disk, and store preload data on your disk.

The script is a revision of this.

Prerequisites

pip install imageio tqdm

Usage

For training data, use parameter mode='train', for validation data, use mode='val'.

from TinyImageNetDataset import TinyImageNetDataset

# for classification
ti = TinyImageNetDataset(root_dir="./path_to_folder", download=True, task='classification')

# for localization
ti = TinyImageNetDataset(root_dir="./path_to_folder", download=True, task='localization')