Improving Deep Learning using Generic Data Augmentation

Abstract

Deep artificial neural networks require a large corpus of training data in order to effectively learn, where collection of such training data is often expensive and laborious. Data augmentation overcomes this issue by artificially inflating the training set with label preserving transformations. Recently there has been extensive use of generic data augmentation to improve Convolutional Neural Network (CNN) task performance. This study benchmarks various popular data augmentation schemes to allow researchers to make informed decisions as to which training methods are most appropriate for their data sets. Various geometric and photometric schemes are evaluated on a coarse-grained data set using a relatively simple CNN. Experimental results, run using 4-fold cross-validation and reported in terms of Top-1 and Top-5 accuracy, indicate that cropping in geometric augmentation significantly increases CNN task performance.

Cited by in corpus (13)

• Towards a Robust Parameterization for Conditioning Facies Models Using Deep Variational Autoencoders and Ensemble Smoother
• Recent Advances in Object Detection in the Age of Deep Convolutional Neural Networks
• Assuring the Machine Learning Lifecycle: Desiderata, Methods, and Challenges
• Recent Developments Combining Ensemble Smoother and Deep Generative Networks for Facies History Matching
• Ensembles of Convolutional Neural Networks models for pediatric pneumonia diagnosis
• Apple Counting using Convolutional Neural Networks
• Deep Semantic Segmentation in an AUV for Online Posidonia Oceanica Meadows identification
• Agile Effort Estimation: Have We Solved the Problem Yet? Insights From A Replication Study
• Latent Causal Invariant Model
• A Preliminary Study on Data Augmentation of Deep Learning for Image Classification
• Controllable Data Augmentation Through Deep Relighting
• Label Denoising with Large Ensembles of Heterogeneous Neural Networks
• C2G-Net: Exploiting Morphological Properties for Image Classification