Random-Erasing icon indicating copy to clipboard operation
Random-Erasing copied to clipboard

Random Erasing Data Augmentation. Experiments on CIFAR10, CIFAR100 and Fashion-MNIST

Results 11 Random-Erasing issues
Sort by recently updated
recently updated
newest added

Hi @zhunzhong07 Thanks for your amazing work. Can you sharing your trained models for CIFAR10, CIFAR100 and Imagenet? I want all these models for my research work. Thanks,

The resnet44/56/110 use BottleNeck,why n = (depth-2) // 6 ? The bottleNeck has three conv, for example the n of resnet44 is (44-2) // 6 = 7, if use bottleNeck,...

Replaced deprecateed volatile keyword with torch no grad. Replaced deprecated async=True with non_blocking=True.

I am doing target detection, I have seen your data enhancement method, and want to use it, how should I use Random-Erasing on my own data set?

The repository does not contain any code for the VOC image aware and object aware augmentation, and the paper does not details the probabilities and hyper parameters used. Is it...

Can you update the fashion-MNIST results in the paper or mention results in readme file? Mentioned results are inconsistent? or share fashion-MNIST dataset, on which you performed the experment? Thanks

Can you share detection and re-identification code? thanks,

Hello, I have a question. If the shapes that are randomly erased can also be diversified, is it more like mimicking a realistic scene?