Deep structured learning
Seminar Title: Deep structured learning
Speaker: Prof Chunhua Shen, School of Computer Science, The University of Adelaide
Time and Date: 1.30 PM – 2.30 PM, 3 July 2015
Venue: CB11.06.408.FEIT Seminar Room
Abstract: Structured output learning concerns the problem of predicting multiple variables that have dependency, with Conditional random field (CRF) as a typical example. It shows great promise in tasks like semantic image segmentation.
Recently, there is mounting evidence that features from deep convolutional neural networks (CNN) set new records for various vision applications. Here I show how we can combine CRFs with deep CNNs to predict complex labels while considering the dependencies between the output variables.
The first application is to learn depth from single monocular images. Compared with depth estimation using multiple images such as stereo depth perception, depth from monocular images is much more challenging. We propose a deep structured learning scheme which learns the unary and pairwise potentials of continuous CRF in a unified deep CNN framework, termed Deep Convolutional Neural Fields. The second model is a new formulation of deep discrete CRFs that learn both unary and pairwise terms using multi-scale fully convolutional neural networks (FCNNs) in an end-to-end fashion, which enables us to model complex spatial relations between image regions. A naive method for training such an approach would rely on direct likelihood maximization of the CRF, but this would require expensive inference at each stochastic gradient decent iteration, rendering the approach computationally unviable.
We propose a novel method for efficient joint training of the deep structured model based on piecewise training. This approximate training method avoids repeated inference, and so is computationally tractable.
We also demonstrate that it yields results that are competitive with the state-of-the-art in semantic segmentation for the PASCAL VOC 2012 dataset.
Chunhua Shen is a Professor at School of Computer Science, University of Adelaide. He is a Project Leader and Chief Investigator at the Australian Research Council Centre of Excellence for Robotic Vision (ACRV), for which he leads the project on machine learning for robotic vision. Before he moved to Adelaide, he was with the computer vision program at NICTA (National ICT Australia), Canberra Research Laboratory for about six years. His research interests are in the intersection of computer vision and statistical machine learning. He studied at Nanjing University, at Australian National University, and received his PhD degree from the University of Adelaide. From 2012 to 2016, he holds an Australian Research Council Future Fellowship.