Semantic Vectorisation: From Bitmaps to Intelligent Representations. The objective of this innovative project is to provide a solution to the open question of representing natural images by semantically rich vector graphics. The challenges are to identify key visual and temporal elements for images and videos, and efficiently decompose the visual data into semantic vector representations that are faithful to original data, compact and editable. The project aims to investigate new bitmap-to-vecto ....Semantic Vectorisation: From Bitmaps to Intelligent Representations. The objective of this innovative project is to provide a solution to the open question of representing natural images by semantically rich vector graphics. The challenges are to identify key visual and temporal elements for images and videos, and efficiently decompose the visual data into semantic vector representations that are faithful to original data, compact and editable. The project aims to investigate new bitmap-to-vector conversion methods. It is expected to develop a framework where semantic labels and hyperlinks can be embedded in visual data automatically. It hopes to pioneer the creation of a web of images where the links are on image/video regions. New image simplification, stylisation, and non-photorealistic rendering methods are expected to be provided.Read moreRead less
Discovery Early Career Researcher Award - Grant ID: DE150101365
Funder
Australian Research Council
Funding Amount
$360,000.00
Summary
In-person tele-presence through hybrid camera networks. This project aims to develop novel theories and algorithms for live capturing of accurate dense 3D models of moving subjects based on hybrid camera networks. The latter consist of a mix of static external red, green, blue plus depth (RGB-D) cameras and a dynamic head-mounted regular camera. The scientific novelties will be dense, non-rigid, and collaborative structure-from-motion theories that maximise the exploitation of such hybrid inform ....In-person tele-presence through hybrid camera networks. This project aims to develop novel theories and algorithms for live capturing of accurate dense 3D models of moving subjects based on hybrid camera networks. The latter consist of a mix of static external red, green, blue plus depth (RGB-D) cameras and a dynamic head-mounted regular camera. The scientific novelties will be dense, non-rigid, and collaborative structure-from-motion theories that maximise the exploitation of such hybrid information, for instance by utilising exact head-pose information. The outcome is a working prototype producing live full-body animations, thus leveraging new applications in the Information Technology industry. Highly strategically relevant examples are given by 3D tele-presence, enhanced tele-operation, robotics, and intelligent transportation systems.Read moreRead less
A general Bayesian multilinear analysis framework for human behaviour recognition. Smart information use is essential for effective video surveillance in order to guard against accidents, fight crime and combat terrorism. In this project advanced probabilistic methods will be applied to visual surveillance information, to warn of impending accidents and to track criminals and terrorists and predict their behaviours.
Discovery Early Career Researcher Award - Grant ID: DE180101438
Funder
Australian Research Council
Funding Amount
$356,446.00
Summary
Multi-view synergistic learning for human behaviour analysis. This project aims to equip machines with a human-likeability to synergistically harness multiple information sources for the purpose of optimal decision-making. This project will produce the next great step for machine intelligence - laying the theoretical foundation for the learning of multiple views and building the next generation of intelligent systems which can accommodate multiple information sources. This research is fundament ....Multi-view synergistic learning for human behaviour analysis. This project aims to equip machines with a human-likeability to synergistically harness multiple information sources for the purpose of optimal decision-making. This project will produce the next great step for machine intelligence - laying the theoretical foundation for the learning of multiple views and building the next generation of intelligent systems which can accommodate multiple information sources. This research is fundamental to the creation of intelligent systems that elegantly tackle varieties of big data. This should benefit science, society, and the economy nationally through applications including autonomous vehicle development, sensor technologies, and human behaviour analysis.Read moreRead less
Nonlinear Transfer Distance Metric Learning for Gleaning Knowledge from the Crowd. This project will develop nonlinear transfer distance metric learning algorithms for training and test samples that are not independent and identically distributed, or from different instance spaces. New theoretical foundations for crowd-sourcing will lead to innovative intelligent systems for such purposes as the NBN, social, and security services, and keep pace with developments in hardware technology. The outco ....Nonlinear Transfer Distance Metric Learning for Gleaning Knowledge from the Crowd. This project will develop nonlinear transfer distance metric learning algorithms for training and test samples that are not independent and identically distributed, or from different instance spaces. New theoretical foundations for crowd-sourcing will lead to innovative intelligent systems for such purposes as the NBN, social, and security services, and keep pace with developments in hardware technology. The outcomes include applications in social networks, the Internet, and climate change, as well as video surveillance to help combat crime and terrorism. The innovative research will significantly benefit Australia’s economy, environment and society, and will maintain Australia's global leading role in the machine learning and computer vision.Read moreRead less
Streaming label learning for leaching knowledge from labels on the fly. This machine intelligence project aims to explore the potential to use and incorporate past knowledge and training to better understand, interpret and develop new concepts. The expected outcomes will provide major technological breakthroughs to benefit science, society, and the economy nationally by laying theoretical foundations for learning labels in a streaming fashion, and building the next generation of intelligent syst ....Streaming label learning for leaching knowledge from labels on the fly. This machine intelligence project aims to explore the potential to use and incorporate past knowledge and training to better understand, interpret and develop new concepts. The expected outcomes will provide major technological breakthroughs to benefit science, society, and the economy nationally by laying theoretical foundations for learning labels in a streaming fashion, and building the next generation of intelligent systems to accommodate environment change in applications about cybercrime, terrorism, and emergence.Read moreRead less
Efficient multi-view video coding with cuboids and base anchored models. This project aims to address current deficiencies in multi-view video coding technology to achieve the ultra-compression efficiency demanded by increasing display resolutions and synchronised viewpoints. The project expects to generate new knowledge, by moving from the current pixel-centric approach to methods that concentrate information common to many view-frames. The project is expected to improve compression of audio-vi ....Efficient multi-view video coding with cuboids and base anchored models. This project aims to address current deficiencies in multi-view video coding technology to achieve the ultra-compression efficiency demanded by increasing display resolutions and synchronised viewpoints. The project expects to generate new knowledge, by moving from the current pixel-centric approach to methods that concentrate information common to many view-frames. The project is expected to improve compression of audio-visual services that are of great interest to international standards bodies and industry, while facilitating free interaction and augmented reality. This project will provide significant benefits to broadcast, entertainment, surveillance and health industries and position Australia as a world leader in this field.Read moreRead less
Automatic Machine Learning with Imperfect Data for Video Analysis . This project aims to propose new algorithms and technologies for constructing an efficient video analysis system, which will be aligned with Australia’s science and research priorities. Specifically, during this project, a novel network structure search method based on auto machine learning will be proposed, an unsupervised domain adaptation algorithm will be developed, and a generative data augmentation method will be construct ....Automatic Machine Learning with Imperfect Data for Video Analysis . This project aims to propose new algorithms and technologies for constructing an efficient video analysis system, which will be aligned with Australia’s science and research priorities. Specifically, during this project, a novel network structure search method based on auto machine learning will be proposed, an unsupervised domain adaptation algorithm will be developed, and a generative data augmentation method will be constructed. All of these will construct a stable and efficient deep neural network, which is able to process large size videos captured from real scenarios in high efficiencies. Various fields, such as health care service and cybersecurity, will benefit hugely from this project.Read moreRead less
Discovery Early Career Researcher Award - Grant ID: DE190100626
Funder
Australian Research Council
Funding Amount
$393,000.00
Summary
Towards data-efficient future action prediction in the wild. This project aims to build state-of-the-art deep learning models to predict future actions in videos. The project expects to produce the next great step for machine intelligence, the potential to explore a handful of labelled examples to better understand, interpret and infer human actions. Expected outcomes of this project lay theoretical foundations for learning future action prediction in the wild scenario and build the next generat ....Towards data-efficient future action prediction in the wild. This project aims to build state-of-the-art deep learning models to predict future actions in videos. The project expects to produce the next great step for machine intelligence, the potential to explore a handful of labelled examples to better understand, interpret and infer human actions. Expected outcomes of this project lay theoretical foundations for learning future action prediction in the wild scenario and build the next generation of intelligent systems to accommodate limited supervision. This should benefit science, society, and the economy nationally through the applications of autonomous vehicles, sensor technologies, and cybersecurity.Read moreRead less
Video plasticity: Scalable video coding with inherently consistent motion. This project aims to improve how video coders represent motion, leading to more efficient motion descriptions and fewer distinct motion fields. The project will develop motion inference algorithms that ensure consistent motion descriptions throughout a group of pictures, allowing seamless integration of scalable video coding, motion compensated temporal filtering and motion compensated frame interpolation operations. The ....Video plasticity: Scalable video coding with inherently consistent motion. This project aims to improve how video coders represent motion, leading to more efficient motion descriptions and fewer distinct motion fields. The project will develop motion inference algorithms that ensure consistent motion descriptions throughout a group of pictures, allowing seamless integration of scalable video coding, motion compensated temporal filtering and motion compensated frame interpolation operations. The project is expected to support an efficient and interactive video browsing experience, largely decoupled from original frame rate and resolution; and deliver practical solutions that can be efficiently implemented on consumer devices.Read moreRead less