Is it possible to get originals HQ images from CIFAR10 dataset?

Question

I'm currently working on my thesis on the neural networks. I'm using the CIFAR10 as a reference dataset. Now I would like to show some example results in my paper. The problem is, that the images in the dataset are 32x32 pixels so it's really hard to recognize something on them when printed on paper.

Is there any way to get hold of the original images with higher resolution?

UPDATE: I'm not asking for image processing algorithm, but for the original images presented in CIFAR-10. I need some higher resolution samples to put in my paper.

Answer 1

I now have the same problem and I just found your question.

It seems that CIFAR was built from labeling the tinyimages dataset, and are kind enough to share the indexing from CIFAR to tinyimages. Now tinyimages contain metadata file with URL of the original images and a toolbox for getting for any image you wish (e.g. those included in the CIFAR index).

So one may write a mat file which does this and share the results...

Answer 2

They're just small:

The CIFAR-10 and CIFAR-100 are labeled subsets of the 80 million tiny images dataset.

You could use Google reverse image search if you're curious.