Deep learning with data augmentation
joppe boekestijn · 2018
Convolutional neural networks (CNNs) learn better when more training data is presented to the network. Data augmentation techniques artificially increase the amount of image data that can be passed to a CNN. This allows the CNNs to extract class-defining features more accurately, resulting in better classification accuracies. In this thesis several data techniques are presented and their performances are compared. Two deep learning architectures are used, GoogLeNet and ResNet, and they are both trained on a plant image dataset ('Tropic10'). The data augmentation techniques used in this paper are rotation, flipping, shifting, cutout, and mix-up. Their performances are compared, as well as the performance of some combination of techniques. This results in 11 different augmentation methods. Both deep learning architectures are configured with either pre-trained weights, trained on the 'ImageNet' dataset, or with randomly initialized weights. Both configurations benefit from data augmentation techniques, in some experiments leading to a 3% increase of classification accuracy. Especially fipping or shifting the images, or combining them, resulted in the best performance.