Data augmentation is an essential part of the training process applied to deep learning models. The motivation is that a robust training process for deep learning models depends on large annotated datasets, which are expensive to be acquired, stored and processed. Therefore a reasonable alternative is to be able to automatically generate new annotated training samples using a process known as data augmentation. The dominant data augmentation approach in the field assumes that new training samples can be obtained via random geometric or appearance transformations applied to annotated training samples, but this is a strong assumption because it is unclear if this is a reliable generative model for producing new training samples. In this paper, we provide a novel Bayesian formulation to data augmentation, where new annotated training points are treated as missing variables and generated based on the distribution learned from the training set. For learning, we introduce a theoretically sound algorithm --- generalised Monte Carlo expectation maximisation, and demonstrate one possible implementation via an extension of the Generative Adversarial Network (GAN). Classification results on MNIST, CIFAR-10 and CIFAR-100 show the better performance of our proposed method compared to the current dominant data augmentation approach mentioned above --- the results also show that our approach produces better classification results than similar GAN models.
Toan Tran (The University of Adelaide)
Trung Pham (The University of Adelaide)
Gustavo Carneiro (The University of Adelaide)
Gustavo Carneiro is an associate professor of the School of Computer Science at the University of Adelaide. He has joined the University of Adelaide as a senior lecturer in 2011, and has become an associate professor in 2015. In 2014, he spent 7 months at the Technical University of Munich as a visiting professor and a Humboldt fellow, collaborating with Prof. Nassir Navab. From 2008 to 2011 Dr. Carneiro was a Marie Curie IIF fellow and a visiting assistant professor at the Instituto Superior Tecnico (Lisbon, Portugal) within the Carnegie Mellon University-Portugal program (CMU-Portugal). From 2006 to 2008, Dr. Carneiro was a research scientist at Siemens Corporate Research in Princeton, USA. In 2005, he was a post-doctoral fellow at the the University of British Columbia with Professor David Lowe and at the University of California San Diego with Professor Nuno Vasconcelos. Dr. Carneiro received his Ph.D. in computer science from the University of Toronto under the supervision of Professor Allan Jepson in 2004. His main research interest are in the fields of computer vision, medical image analysis and machine learning.
Lyle Palmer (The University of Adelaide)
Ian Reid (University of Adelaide)
More from the Same Authors
2019 Poster: Social-BiGAT: Multimodal Trajectory Forecasting using Bicycle-GAN and Graph Attention Networks »
Vineet Kosaraju · Amir Sadeghian · Roberto Martín-Martín · Ian Reid · Hamid Rezatofighi · Silvio Savarese
2019 Poster: Unsupervised Scale-consistent Depth and Ego-motion Learning from Monocular Video »
Jiawang Bian · Zhichao Li · Naiyan Wang · Huangying Zhan · Chunhua Shen · Ming-Ming Cheng · Ian Reid
2017 Poster: Deep Subspace Clustering Networks »
Pan Ji · Tong Zhang · Hongdong Li · Mathieu Salzmann · Ian Reid
2015 Poster: Deeply Learning the Messages in Message Passing Inference »
Guosheng Lin · Chunhua Shen · Ian Reid · Anton van den Hengel