AlexNet's first form of data augmentation is
ARandom Gaussian noise added to each pixel
BRandom translations and horizontal reflections
CRandom rotations by 90 degree increments only
DRandom colour-jittering by uniform offsets
Answer & Solution
Correct answer: B. Random translations and horizontal reflections
AlexNet extracts random 224x224 patches and their horizontal reflections from 256x256 source images. This enlarges the training set by a factor of ~2048.
Related questions
Image similarity in AlexNet's last hidden layer is measured byThe SGD update rule for AlexNet usesRemoving any single middle convolutional layer of AlexNet costs roughlyAlexNet was trained for roughly how many epochs on ImageNet?At test time dropout is approximated byAlexNet uses overlapping pooling withAlexNet's second form of data augmentation isWhy are ReLU networks faster to train than tanh networks?