Cross Domain Dataset

[sylvialee12]

I'm wondering if this method can be applied to datasets containing different domains such as dataset of ImageNet, which we have cats, flowers, dog, etc?

[tobschu94]

@sylvialee12 I wanted to try something similiar, I have a dataset with pedestrians in cities and nature and wanted to augment these images. Have you any experience in using such complex data?

[mohsin127]

Hey
I also have some random data and want to augment it through DAGAN. Does anyone of you have any experience with that?

[lerndeep]

could someone help me to prepare .nzp file for a custom dataset?

[uzaymacar]

I am running data augmentation experiments in the CIFAR10 dataset, but unable to get any good results as of yet. The easiest way to understand the data format is by loading the .npy mentioned in the README under Datasets. It has a shape of (num_classes, num_images, height, width, num_channels). This asks that each of your classes cpntains the same number of images, which is true for the Omniglot dataset but not might not be true for other datasets. In the latter cases, you can append (np.append()) the NumPy array with itself as many times as needed to increase num_images to the maximum num. of images you have for any class.

[jerofad]

@uzaymacar would you mind sharing a piece of your code on how you made it work for CIFAR10? I am also having troubles with this regards and I cannot quite understand how you did yours.