Image Augmentation with Convolutional Neural Networks

Data Augmentation is a data-space solution to the problem of limited data. Data augmentation uses a suite of techniques to enhance the size and quality of training datasets in order to build up better Deep Learning models. Convolutional Neural Networks (CNNs) have achieved significant improvements in computer vision due to the efficient computing power and large volumes of images. However, image collection is an expensive and challenging process. It is particularly challenging to obtain satisfactory performance with a limited dataset due to overfitting arising in the training process. To address this issue, a convolutional neural network model is built up with image augmentation to combat overfitting in this study. The available image data is augmented by implementing blending modes to expose the full tonal range contained in each training image. Additionally, a horizontal flip transformation to create mirror versions of training images is implemented. These data augmentations are shown to reduce overfitting in CNN models. Moreover, different combinations of blending mode layers are explored to maximize the validation accuracy of the network model.