Awesome
Downloads Cifar datasets directly from http://www.cs.toronto.edu/~kriz/cifar.html and converts them to Torch tables.
Cifar-100 format
Writes two files: cifar100-train.t7, cifar100-test.t7 Each of them is a table of the form:
th> c100 = torch.load('cifar100-train.t7')
th> print(c100)
{
labelCoarse : ByteTensor - size: 50000
data : ByteTensor - size: 50000x3x32x32
label : ByteTensor - size: 50000
}
Cifar-10 format
Writes two files: cifar10-train.t7, cifar10-test.t7 Each of them is a table of the form:
th> c10 = torch.load('cifar10-train.t7')
th> print(c10)
{
data : ByteTensor - size: 50000x3x32x32
label : ByteTensor - size: 50000
}