PyTorch custom dataset APIs -- CUB-200-2011, Stanford Dogs, Stanford Cars, FGVC Aircraft, NABirds, Tiny ImageNet, iNaturalist2017
This repo contains some unofficial PyTorch dataset APIs(mainly for Fine-Grained Visual Categorization task), which support automatically download (except large-scale datasets), extract the archives, and prepare the data.
The code was tested on
Use them the similar way you use torchvision.datasets
.
train_dataset = Cub2011('./cub2011', train=True, download=False)
test_dataset = Cub2011('./cub2011', train=False, download=False)
Feel free to open an issue or PR.
MIT