Kodak24#
- class deepinv.datasets.Kodak24(root=None, download=False, transform=None, verbose=True, use_dict_output=False)[source]#
Bases:
ImageFolderDataset for Kodak24.
The Kodak24 dataset [1] is a dataset consisting of 24 images commonly used for testing performance of image reconstruction algorithms. Images have a fixed size of 768Γ512 (or 512Γ768) pixels.
Raw data file structure:
self.root --- Kodak-Lossless-True-Color-Image-Suite-master.zip | --- Kodak-Lossless-True-Color-Image-Suite-master --- PhotoCD_PCD0992 --- 01.png | | | --- 02.png | --- ... | --- xxx
Raw dataset source : MohamedBakrAli/Kodak-Lossless-True-Color-Image-Suite
- Parameters:
root (str) β Root directory of dataset. Directory path from where we load and save the dataset.
download (bool) β If
True, downloads the dataset from the internet and puts it in root directory. If dataset is already downloaded, it is not downloaded again. Default at False.transform: (Callable) β (optional) A function/transform that takes in a PIL image and returns a transformed version. E.g,
torchvision.transforms.RandomCropverbose (bool) β Print a message if the dataset has been correctly downloaded. Default
True.use_dict_output (bool) β whether to return output as dict with keys βxβ, βyβ, βparamsβ instead of tuple (default
False).
- References: