paper_reimplementation

I followed-up berniwal/swin-transformer-pytorch for swin_transformer codes

Randaugment: Practical automated data augmentation with a reduced search space

• Features Applies randomly selected augmentations with fixed magnitude to each mini-batch Supports PIL image as input under torchvision version 0.8.0 Supports PIL image and torch tensor under torchvision version 0.9.0

• Usage

import torch
import torchvision
import torchvision.transform as tt
from randaugment import RandAugment

RA = RandAugment(3, 3)  # N, M
temp = tt.functional.to_pil_image(torch.randn(3, 5, 5))
result1 = RA(temp)
result2 = RA(temp)

dataset = torchvision.datasets.CIFAR10(..., transform=tt.Compose([RA]), ...)