cifar10_make_original.lua

require 'xlua'
require 'sys'

local batches_folder = '/opt/rocks/cifar.torch/cifar-10-batches-t7'

local data = {}
local labels = {}

for i=1,5 do
   local name = paths.concat(batches_folder, 'data_batch_'..i..'.t7')
   local part = torch.load(paths.concat(batches_folder, name), 'ascii')
   table.insert(data, part.data:view(3,32,32,-1))
   table.insert(labels, part.labels:squeeze())
end

data = torch.ByteTensor.cat(data, 4)
labels = torch.ByteTensor.cat(labels)

test_part = torch.load(paths.concat(batches_folder, 'test_batch.t7'), 'ascii')
test_labels = test_part.labels
test_data = test_part.data


local dataset = {
   trainData = {
      data = data:permute(4,1,2,3):clone(),
      labels = labels:add(1),
      size = function() return labels:numel() end,
   },
   testData = {
      data = test_data:view(3,32,32,-1):permute(4,1,2,3):clone(),
      labels = test_labels:squeeze():add(1),
      size = function() return test_labels:numel() end,
   }
}

print(dataset)
print(dataset.trainData.labels:max())
print(dataset.testData.labels:max())
torch.save('cifar10_original.t7', dataset)

Thanks sharing your code. How many training samples per epoch after data augmentation as mentioned in Experimental Results section?
https://arxiv.org/pdf/1605.07146.pdf

For experiments we chose well-known CIFAR-10, CIFAR-100, SVHN and ImageNet image
classification datasets. CIFAR-10 and CIFAR-100 datasets [17] consist of 32 × 32 color
images drawn from 10 and 100 classes split into 50,000 train and 10,000 test images. For data
augmentation we do horizontal flips and take random crops from image padded by 4 pixels on
each side, filling missing pixels with reflections of original image