Load CIFAR-10 in NDArrays

loadCifar10NDA.py

import mxnet as mx

def buildTrainingSet(path):
    training_data = []
    training_label = []
    for f in ("data_batch_1", "data_batch_2", "data_batch_3", "data_batch_4", "data_batch_5"):
        imgarray, lblarray = extractImagesAndLabels(path, f)
        if not training_data:
            training_data = imgarray
            training_label = lblarray
        else:
            training_data = mx.nd.concatenate([training_data, imgarray])
            training_label = mx.nd.concatenate([training_label, lblarray])
    return training_data, training_label


path="cifar-10-batches-py/"
batch=128

training_data, training_label = buildTrainingSet(path)
train_iter = mx.io.NDArrayIter(
    data=training_data, label=training_label, batch_size=batch, shuffle=True)

valid_data, valid_label = extractImagesAndLabels(path, "test_batch")
valid_iter = mx.io.NDArrayIter(
    data=valid_data, label=valid_label, batch_size=batch, shuffle=True)

print training_data.shape
print training_label.shape
print valid_data.shape
print valid_label.shape