PhD Student @ AIML
hayden.faulkner _at_ adelaide.edu.au
I am a final year PhD student within The Australian Institute of Machine Learning at The University of Adelaide, in South Australia. I work heavily in the fields of Computer Vision, Machine Learning and Artificial Intelligence. I am currently focused on developing novel approaches to automatically understand and summarise videos. I have worked on image and video classification, detection, captioning and tracking with deep neural networks and big data. I am passionate about applying the latest techniques to solve real world problems, as well as effectively communicating ideas to both collegues and the broader community.
Using computers to perform image and video processing. Extracting useful information from image data.
Enabling computers to automatically learn from data to uncover trends and extract information otherwise unseen.
Experience building and utilising many forms of neural networks, such as convolutional and recurrent neural networks.
Ability to turn the latest state of the art techniques into useful end user applications to solve real world problems.
This paper introduces a new video understanding dataset which can be utilised for the related problems of event recognition, localisation and description in video. Our dataset consists of dense, well structured event annotations in untrimmed video of tennis matches. We also include highly detailed commentary style descriptions, which are heavily dependent on both the occurrence as well as the sequence of particular events. We use general deep learning techniques to acquire some initial baseline results on our dataset, without the need for explicit domain-specific assumptions.
This paper is an empirical study of the application of visual detection and tracking methods to the problem of locating and tracking all AFL players during a game. While most person detection and tracking algorithms are designed for pedestrians, we show that with appropriate modifications, state of the art methods can be adapted to a more challenging domain where motion is significantly more varied and occurs in a much wider area.
We analyse experimentally the region covariance descriptor which has proven useful in numerous computer vision applications. The properties of the descriptor–despite its widespread deployment–are not well understood or documented. In an attempt to uncover key attributes of the descriptor, we characterise the interdependence between the choice of features and distance measures through a series of meticulously designed and performed experiments. Our results paint a rather complex picture and underscore the necessity for more extensive empirical and theoretical work. In light of our findings, there is reason to believe that the region covariance descriptor will prove useful for methods that perform image super-resolution, deblurring, and denoising based on matching and retrieval of image patches from an image dictionary.
This study addresses the recently introduced Traveling Thief Problem (TTP) which combines the classical Traveling Salesman Problem (TSP) with the 0-1 Knapsack Problem (KP). The problem consists of a set of cities, each containing a set of available items with weights and profits. It involves searching for a permutation of the cities to visit and a decision on items to pick. A selected item contributes its profit to the overall profit at the price of higher transportation cost incurred by its weight. The objective is to maximize the resulting profit. We propose a number of problem-specific packing strategies run on top of TSP solutions derived by the Chained Lin-Kernighan heuristic. The investigations provided on the set of benchmark instances prove their rapidity and efficiency when compared with an approximate mixed integer programming based approach and state-of-the-art heuristic solutions from the literature.
2010 - 2012
GPA of 5.625