computer_vision_datasets's Introduction

Computer Vision Datasets

datasetninja.com has an impressive collection of comptuer vision datasets. This package was built to let you easily download and use these computer vision datasets.

Installation

pip install git+https://github.com/franchesoni/computer_vision_datasets.git or add the only relevant file in this repo (computer_vision_datasets/module.py) to your codebase.

Available datasets (code)

To check the available datasets run list-ninja or

from computer_vision_datasets import get_released_datasets
print(sorted(get_released_datasets().keys()))

Download a dataset

Find the name of the dataset you want with the code above. Here we use 'ADE20K'.

from computer_vision_datasets import download
download('ADE20K', '/your/destination/directory')

The `SegDataset` class

There is a SegDataset class for segmentation datasets. You just need to point it to the dataset path. Example:

ds = SegDataset(ds_path, split='test')

You can, for instance, wrap it in a Pytorch Dataset:

from torch.utils.data import Dataset

class PyTorchWrapperDataset(Dataset):
    def __init__(self, ds):
        # super().__init__() is not needed since we're not overriding anything from the parent's __init__
        self.original_dataset = ds

    def __getitem__(self, index):
        # Assuming the original dataset uses __getitem__ to access items
        # If it uses a different method (like get_item), adjust accordingly.
        return self.original_dataset[index]

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

Recommend Projects