Project offerings for Semester 2, 2017

Click on the supervisor's name for a description of the projects they are offering.

Projects will be added over the coming weeks.


Supervisor Project Credit points
Vera Chung   

 

 

 

 

    

 

 

 

 

 

 

 

 

New object tracking algorithms for 3D videos 12 or 18

A new deep learning based model for Lightfield image classification

12 or 18

Mobile Application for Virtual Reality (VR) Contents Capture, Sharing, and Interaction

18

Virtual Reality (VR) Multimedia System based on Light Field Technology

18

Augmented Reality based on Constructed 3D Scene 

18
Azadeh Ghari Neiat and Athman Bouguettaya

Designing  a  failure-proof  model  for  crowdsourced  sensor-cloud  service  selection  and  composition

12
Vincent Gramoli   

 

 

Secure Red Belly Blockchain Wallet

18

Secure Blockchain Explorer

18

Blockchain exchange platform

18
Seokhee Hong      

Scalable Visual Analytics

12 or 18

Visualisation and Analysis of Massive Complex Social Networks and Biological Networks

12 or 18

Navigation and Interaction Techniques for 2.5D Network Visualisation

12 or 18
Tongliang Liu     

Modelling label noise in big data

12 or 18
David Lowe         

 

 

 

 

 

 

Lab augmentation 2 : Development of a laboratory augmentation prototype that demonstrates the feasibility of using current mobile phones to support augmentation of standard laboratory experiments

12 or 18

MOOLS: Massive Open Online Labs

12 or 18

Using virtual reality augmentation to support simultaneous use of physical equipment

12 or 18

Enhancing laboratory learning through scripted guidance using Smart Sparrow

12 or 18
Sajib Mistry and Athman Bouguettaya

Developing a Deep Learning Platform for Social Cloud Services

12
Quan Nguyen 

 

 

Quality Aspects of Big Graph Visualization

12 or 18

Methods for Big Graph Visualization

12 or 18
Lie Qu and Athman Bouguettaya

Detecting Junk Services based on Machine Learning in SOA Environments

12
Bernhard Scholz     

 

 

 

 

 

Matching Service (with RateSetter Australia)

12

An Integrated Development Environment for the Cloud

12

Performance Benchmark Suite for Logic Oriented Programs

12

Python Language Bindings for Soufflé

12

Debugger for Soufflé

12

Smart Contracts: Blockchain Security

12

 Zhiyong Wang

 

 

 

    

Multimedia Data Summarization

12 or 18

Human Motion Analysis, Modeling, Animation, and Synthesis

12 or 18

Predictive Analytics of Big Time Series Data

12 or 18
Remote-sensing Image Analysis 12 or 18

Audio Data Analysis

12 or 18
Chang Xu   

Recommendation with diversified interactions between customers and products

18

Learning Smaller Deep Neural Networks for Image Classification

18
Industry projects - to express interest in these projects please email your CV to Evelyn Riegler (evelyn.riegler@sydney.edu.au)              

 

 

 

           

Multi User Clinical Holography (MUCH)

12

The Trend Report

12

A markup language authoring platform for the production of learning resources

12

Development of a workflow program to validate compounding techniques to make medicines using the HoloLens®

12
Development of a workflow program for real-time in-clinic visualization of paediatric cardiac MRIs in Oculus Rift 12
Medication and Prescription Cross-Checking Optimisation 12
Sharing Haswell’s educational legacy and the University’s scientific heritage 12
Quality Assurance System Using Machine Learning & Computer Vision 12
Identification of common cluster of diseases associated with primary diagnosis of a heart condition and analysis of variations in costs and length of stay for patients with different clusters of diseases 12
Study of variation in length of stay and cost of care over time, for high cost surgical procedures for patients with different health profiles 12
Online Authorized Entry Signage generator
ALSO populates Hazardous Areas register
12
Our Pills Talk Medication Safety App 12

 

Project supervised by Vera Chung

New Object Tracking algorithms for 3D Videos (12cp or 18cp)
Object Tracking has been an important research topic due to a large number of application area including human computer interaction, human motion analysis and surveillance. Most object tracking research is conducted on videos captured from RGB cameras. There is a recent surge in the availability of RGBD sensors that produces image and depth data. This project aims to develop an object tracker for tracking objects in RGBD 3D videos.

Requirements: Good programming in Python and Mathlab

A new Deep Learning Based model for Lightfield Image Classification (12cp or 18cp)
Lightfield photography is recently getting popular due to the development of handheld lightfield cameras by company like Lytro. The capture of lightfield enables the reconstruction of images with simulated depth of focus at any point on the image. The provision of open sourced lightfield datasets enables building of learning based models on topics such as lightfield image super-resolution, classification and depth estimation. This project aims to develop a new deep learning based model to classify lightfield images.

Requirements: Good programming in Python and Mathlab

Mobile Application for Virtual Reality (VR) Contents Capture, Sharing, and Interaction (18cp)
In this project, you will develop a mobile application that allows a user to use single mobile to take multiple images and stitch them into a VR image for interaction and display. The stitched VR image is expected to be visually pleasant and the stitching speed will be fast. After generating the VR image, the user can interact with the VR image and add clickable labels, annotations, etc. according to their needs. Then, the user can share the VR image in social networks at different clients including the same mobile app, web clients, or head mounted displays like Oculus Rift.

Requirements: This project requires good programming skills. Proven iOS programming experience is a plus.

Virtual Reality (VR) Multimedia System based on Light Field Technology (18cp)
The user experience of current VR multimedia system is not realistic and quite limited due to the media capture and processing technologies. Light field technology can capture the amounts and the directions of all the lights reflected from an object, thus it is able to offer very realistic and immersive media experience especially in a VR environment. This project will study the light field image processing and interaction techniques and incorporate them into a light field based VR multimedia system.

Requirements: This project requires good programming skills. Proven experience with image processing, machine learning, or Unity3D programming is a plus.

Augmented Reality based on Constructed 3D Scene (18cp)
This project will study the technology in naturally augmenting virtual characters or objects in a 3D scene reconstructed from captured images of stereo cameras. The reconstructed 3D scene and the depth map will be provided directly. The scene will be a far-field scene that may cover 100 meters in distance, e.g. a view in a park with the depths of all the real objects in this view are known. The virtual characters and objects are augmented in the scene, animated and naturally mixed in the scene with consideration of the depths of other real objects in the scene.

Requirements: This project requires good programming skills. Proven experience with augmented reality is a plus.

Project supervised by Azadeh Ghari Neiat and Athman Bouguettaya

The contact person for this project is Sajib Mistry.

Designing a failure-proof model for crowdsourced sensor-cloud service selection and composition (12cp)
The aim of this project is to develop a failure-proof model for crowdsourced journey service selection and composition. Since the failures may occur as the user travels on an optimal path, the availability of a previously selected journey service (e.g. bus, train or tram) may turn false and as a result, the travel plan has now formally failed. In such case, the initial travel plan needs to be replanned to deal with the exception. We will investigate state-of-art failure-proof travel planning algorithms to deal with the failure of journey services. First, we will focus on developing a spatio-temporal graph which models the changes in a public transport network (i.e., spatio-temporal network). The graph keeps the track of changes through collecting sensor data from either developed sensors on buses, trams and trains or smartphones which enable commuters (i.e., the crowd) to contribute as multi-modal sensors. The model considers these changes in the network topology in which the existing nodes and edges can be removed and new nodes and edges can be added. We will then design supervised or unsupervised approaches including Hidden Markov Model (HMM), Artificial Neural Networks (ANN) and Autoregressive Integrated Moving Average (ARIMA) for arrival and departure time prediction with or without prior knowledge. Finally, we will explore the use of incremental heuristic search methods (e.g., incremental replanning algorithms such as D*-Lite) to repair the initial travel plan when new information about the environment is received. The proposed approach continually improves its initial travel plan and find the best travel plan from a given source-point to a given destination point while failures happen. The project is a part of a research project about spatio-temporal composition of sensor-cloud services in which commuters use the sensor-cloud services to find the most optimal travel plan. This project will also provide information to students on the implications and additional steps required to model it.

Skill Required:
1. Object-Oriented Programing (OOP) experience in Python or Java.
2. Basic knowledge in Machine Learning.

Project supervised by Vincent Gramoli

Secure Red Belly Blockchain Wallet (18cp)
Building upon recent research achievements in distributed computing, the University of Sydney has just developed the fastest blockchain, the Red Belly Blockchain.

Unfortunately, at this stage there is no secure wallet application, and only expert users can make use of the blockchain to transfer usyd-coin.
In this project, you will join the Concurrent Systems Research Group to design a new safe wallet that can tolerate up to a configurable amount t of hackers.

The wallet will allow a user to retrieve the balance of his address (i.e., account), to generate new public-private key pairs and to send usyd-coins to his contact list.

To this end, the wallet will contact between t+1 and 2t+1 of the peers running the Red Belly Blockchain and identify whether some respondents were hikacked and if it is necessary to send the request to other servers of if the response is actually correct.
This project requires knowledge in mobile application development and excellent programming skills in Java or Swift.

More information

Secure Blockchain Explorer (18cp)
Building upon recent research achievements in distributed computing, the University of Sydney has just developed the fastest blockchain, the Red Belly Blockchain.

Unfortunately, at this stage there is no way to simply observe the coins transferred through the blockchain, and only expert users can guarantee the security of the stored information.

In this project, you will join the Concurrent Systems Research Group to design a transparent and secure storage for the Red Belly Blockchain that will display real-time statistics onto a web server interface.
The research challenge consists of guaranteeing the integrity of the information displayed despite the presence of potential attacks against t of the storage servers.

The replication of the storage will be used to guarantee the persistence of a correct copy of the blockchain and Merkle trees will be used to detect the integrity violation on the copies of the blochchain.

This project requires excellent programming skills and knowledge of network and/or distributed systems.

More information.

Blockchain exchange platform (18cp)
Building upon recent research achievements in distributed computing, the University of Sydney has just developed the fastest blockchain, the Red Belly Blockchain.

In this project, you will join the Concurrent Systems Research Group to design an efficient blockchain exchange platform.

The goal of this project is to design a web platform that consists of monitoring the balance of multiple accounts and crediting new coins.

It will allow to connect with existing payment platforms and generate coins on the user account based on the amount transferred from fiat currency in a simulated environment.

The research challenge will be to provide a platform that will guarantee the security and legitimacy of the transfers through KYC and authentication.

This project requires excellent programming skills and knowledge of electronic payment systems.


More information.

Project supervised by Seokhee Hong

Scalable Visual Analytics (12 or 18cp)
Technological advances such as sensors have increased data volumes in the last few years, and now we are experiencing a “data deluge” in which data is produced much faster than it can be used by humans.
Further, these huge and complex data sets have grown in importance due to factors such as international terrorism, the success of genomics, increasingly complex software systems, and widespread fraud on stock markets.

We aim to develop new visual representation, visualization and interaction methods for humans to find patterns in huge abstract data sets, especially network data sets.

These data sets include social networks, telephone call networks, biological networks, physical computer networks, stock buy-sell networks, and transport networks.

These new visualization and interaction methods are in high demand by industry.

Visualisation and Analysis of Massive Complex Social Networks and Biological Networks (12 or 18cp)
Recent technological advances have led to many massive complex network models in many domains, including social networks, biological networks, webgraphs and software engineering.

Visualization can be an effective analysis tool for such networks. Good visualisation reveals the hidden structure of the networks and amplifies human understanding, thus leading to new insights, new findings and possible prediction of the future.

However, visualisation of such massive complex networks is very challenging due to the scalability and the visual complexity.

This project addresses the challenging issues for visualisation and analysis of massive complex networks by designing and evaluating new efficient and effective algorithms for massive complex social networks and biological networks.

In particular, integration of good analysis method with good visualisation method will be the key approach to solve the research challenge.

Navigation and Interaction Techniques for 2.5D Network Visualisation (12 or 18cp)
Recent technological advances have led to many large and complex network models in many domains, including social networks, biological networks, webgraphs and software engineering.

Visualization can be an effective analysis tool for such networks; good visualisation may reveal the hidden structure of the networks and amplifies human understanding, thus leading to new insights, new findings and possible prediction of the future.

However, visualisation itself cannot serve as an effective and efficient analysis tool for large and complex networks, if it is not equipped with suitable interaction and navigation methods.

Well designed and easy-to-use navigation and interaction techniques can enable the users to communicate with visualization much faster and effectively to perform various analysis tasks such as finding patterns, trends and unexpected events.

Recently, 2.5D graph visualization methods have been successfully applied for visualization of large and complex networks, arising from biological networks, social networks and internet networks.
However, the corresponding navigation method has yet been investigated so far.

This project aim to design, implement and evaluate new methods for navigating 2.5D layouts of large and complex networks to enable users to perform analytical tasks.

Project supervised by Tongliang Liu

Modelling label noise in big data (12 or 18cp)
As data grows bigger and bigger, correctly labelling them is becoming troublesome, expensive, and time consuming. Cheap and fast ways for providing labels, e.g., crowdsourcing, are therefore proposed and becoming popular at the price that the obtained labels are at some level erroneous. The label noise will harm the performances of many modern machine learning algorithms. It is of great important to model the label noise and reduce its influence in the learning procedure.

The research of label noise is at its early stage. The simple random label noise model, which assumes that each label has a probability to randomly flip to a wrong label, is still under intensive study. Problems such as how to efficiently and effectively estimate the flip rate are still remained elusive. In this project, the student will work with me and the research team to find some applications where label noise exists; or to further study the random label noise problem; or to build new insightful label noise models to interpret the real-world data.

This project will equip participants with strong skills in machine learning, but requires them to have some background in applied mathematics and statistics.

Project supervised by David Lowe

Lab augmentation 2 : Development of a laboratory augmentation prototype that demonstrates the feasibility of using current mobile phones to support augmentation of standard laboratory experiments (12 or 18cp)
The outcome will be a prototype phone app that allows students to point their phone camera at a set of laboratory apparatus and have it supplement the display with additional information related to the apparatus. (Extension) The augmented information varies depending on additional information retrieved live from an external source (nominally connected to the equipment, so that the information represents the current state of the apparatus.

MOOLS: Massive Open Online Labs (12 or 18cp)
This project involves investigation of strategies that allow multiple users to share control of a single item of physical laboratory equipment, with the objective of allowing each user to feel as though they are an active participant in the resultant behaviour of the equipment. The core outcome will be development of the software interface for an online heat transfer experiment that allows gamified shared control of a set of laboratory equipment.

Using virtual reality augmentation to support simultaneous use of physical equipment (12 or 18cp)
This project will adapt concepts from earlier work on the use of augmentation of an experimental environment to allow multiple users to simultaneously undertake experimentation on the same item of laboratory equipment. The equipment will be designed to allow each user will have their own virtual software agent (manifested just to them using augmented reality) which reacts to the behaviour of the equipment. The outcome will be a simple prototype that demonstrates the feasibility of the approach.

Enhancing laboratory learning through scripted guidance using Smart Sparrow (12 or 18cp)
Investigation of the feasibility of using Smart Sparrow to provide adaptive guidance in carrying out a physical laboratory experiment. This will require consideration of the ways in which the Smart Sparrow adaptation engine can respond to events drawn from the real world (and in particular from the equipment under exploration). The outcome will be an implementation and evaluation of a proof-of-concept prototype and a set of recommendations regarding feasibility and possible design issues.

Project supervised by Sajib Mistry and Athman Bouguettaya

Developing a Deep Learning Platform for Social Cloud Services (12cp)
Our objective is to transform social sensor data into social cloud services. The physical sensors (e.g. surveillance cameras, GPS and mobile phones) and social sensors (i.e. posts in social media platforms) provide an opportunity to design innovative sensor and cloud applications delivered as services. It is a fairly recent paradigm where services are treated as first-class objects which constitute computing proxies for a wide range of artefacts, such as applications and data. Transforming social media data into social cloud services enable us to compose services based on functional and non-functional (e.g. space and time) requirements. For example, social sensor services can be composed to detect various events such as accidents, traffic congestions and the spread of diseases without a costly infrastructure setup. Subjective metrics in the social cloud can be used to evaluate financial market performances without the costly objective metrics. We will investigate state-of-art technologies to transform social media data into social cloud services. Primarily, we focus on Deep Learning technologies to create social cloud services. We will design the machine learning platform as a foundation to conduct future experiments in research related to social cloud services. The design may involve the following tasks.

1. Collecting data from various social media platforms e.g. Twitter, Facebook, YouTube and EBay.
2. Modifying Theano (a Deep Learning Platform) for the social media dataset.
3. Creating a classifier for the Deep Learning in sensor cloud services.

Supervisor: Sajib Mistry and Prof. Athman Bouguettaya

Skill Required:
1. Object-Oriented Programing (OOP) experience in Python.
2. Basic knowledge in Artificial Neural Networks (ANN) and Deep Learning.

Projects supervised by Quan Nguyen

Methods for Big Graph Visualization (12 or 18cp)
Big graphs are omnipresent in the real-world. Examples include the world wide web, social media, emails, financial records, software systems and system logs. This project’s goal is to develop new visualization methods and tools to help derive more insights from the data.

Specific requirements/skills (if applicable): English language competency; Experience in algorithm design and implementation; Experience in C++, Java or Javascript programming languages; Experience working with Linux and / or Windows Operating Systems.

Quality Aspects of Big Graph Visualization (12 or 18cp)
Very large graphs can be seen from many domains including biological networks, social networks and computer networks. Visualizations give us insights of the data and enable more effective task performance and decision making. This project aims to investigate the quality of the current state-of-the-art tools and techniques for visualizations of large graphs.

Specific requirements/skills (if applicable): English language competency; Experience in algorithm design and implementation; Experience in C++, Java or Javascript programming languages; Experience working with Linux and / or Windows Operating Systems.

Project supervised by Lie Qu and Athman Bouguettaya

The contact person for this project is Sajib Mistry.

Detecting Junk Services based on Machine Learning in SOA Environments (12cp)
Service-oriented Architecture (SOA) has become the most popular paradigm of solution design, where individual services are offered as components for a whole system. Before consuming a component service, it is quite necessary to identify whether it would perform well. However, it is usually a greatly challenging issue when no sufficient historical records of component services exist. Hence, this project addresses the problem of detecting junk services based on machine learning algorithms via their technical specifications, textual descriptions, etc. For example, when a work requester posts a task in a crowdsourcing platform, such as Freelancer, the platform wants to know not only which workers can provide suitable services, but also figure out whether the requester posts an achievable and reasonable task. To this end, workers and the requester’s basic information and the specific information of the task should be collected and analyzed first, and then a detection approach should be designed to take these information as input, and output the success probability of the task. This project focuses on studying the performance of machine learning techniques on this problem. Any type of learning techniques (including deep learning) may be tried and tested in the project in order to find out the best solution. The main tasks involved in this project may contain:
1. Applying web API or crawler to collect data from SOA platforms, such as Freelancer and ProgrammableWeb
2. Cleaning and analyzing the collected data
3. Creating detectors through machine learning toolkits and framework, such as scikit-learn and tensorflow.

Skill Required:
1. Object-Oriented Programing (OOP) skills in Python
2. Basic knowledge in machine learning
3. Ability of technical document reading and self-learning

Projects supervised by Bernhard Scholz

Matching Service (12cp)
(in collaboration with RateSetter Australia)
RateSetter has borrowers and lenders that need to be matched together in a market place. If a match is made the lender will have loaned their money to the borrower. A borrower market order will contain a rate, amount and various other attributes such as borrowing term, purpose of loan, risk category. Lender market orders will contain a rate, amount and rules to target specific attributes of borrower orders.

In order for this market to be flexible, the rules defined on a lender order can be an arbitrarily complex predicate. An example being:
And(EqualTo(LoanPurpose, Business), Or(GreaterThan(Business.Age, 2), EqualTo(Business.Age, 2)))

The borrower order ‘profile’ can contain arbitrary information such as: { Name: {First: ‘John’, Last: ‘Smith’}, Purpose: ‘Car’, Business: {Age: ‘5’}}. The core of this problem is to match lender orders against borrower orders and vice versa based on these predicates.

It is trivial to evaluate lender predicates against borrower orders in a naïve way, however when there are hundreds of thousands of existing lender orders and a new borrower order arrives, the algorithm/data structures must be capable of finding all matches in real time.

Deliverables:

  • Solution design covering algorithms / data structures
  • Implementation of design
  • Performance results

Industry Supervisor: Kym McGain

An Integrated Development Environment for the Cloud (12cp)
Integrated Development Environments (IDEs) are essential for modern software engineering. They accelerate the development cycle from the source code to the binary, and facilitate debugging. However, IDEs are normally executed on a personal desktop or laptop computer.

This project investigates how standard technologies can be fused to a simple IDE in the cloud. The standard technologies comprise a remote UNIX shell, a remote editor, and some visualisation techniques using HTML, Javascript, and web services.

Required Skills: web technologies, HTML, Javascript

Performance Benchmark Suite for Logic Oriented Programs (12cp)
The performance of logic oriented programs is not well understood. To obtain a deep insight of common runtime behaviour of logic programs, benchmarks are required to test various aspects of the runtime behaviour. To collate, archive and present benchmark results in a systematic way, a benchmark harness is required. The benchmark harness is the driver for the benchmark execution.

This project designs and implements a benchmark harness for logic-oriented programming engines. The harness should be written in a scripting language such as Python. The presentation of the benchmark results should be performed with web technologies such as HTML, and Javascript.

Required Skills: scripting language such as python

Python Language Bindings for Soufflé (12cp)
Soufflé is an open-source translator for a logic-oriented programming language. The translator compiles a declarative specification into parallel C++ code. Soufflé's applications include security specifications for SDNs, and checking security properties of large-scale Java programs. Soufflé can be fully integrated with other languages. It has currently a JNI and a C++ interface. To enhance the interoperability of Soufflé, we would like to have interfaces for Python.

In this project, we implement a language binding for Python. The challenges of this project will be to replicate relational data-structures in the Python language and implement high-performance interfaces such that information can be exchanged between the Python language and Soufflé efficiently.

Requirements: some Python and C/ C++ knowledge

Debugger for Soufflé (12cp)
Soufflé is an open-source translator for a logic-oriented programming language. The translator compiles a declarative specification into parallel C++ code. Soufflé's applications include security specifications for SDNs, and checking security properties of large-scale Java programs. Soufflé has no dynamic query interface for querying tuples.

The aim of this project is to implement a simple query language that can retrieve information from the computed logical relations. The query language

Requirements: good C++ knowledge

Smart Contracts: Blockchain Security (12cp)
Smart contracts are computer programs that enforce a set of rules and work seamlessly in conjunction with the blockchain. Although it was mentioned on occasion in the early 2000s, the concept of smart contracts has been recently popularised by the Ethereum programmer Vitalik Buterin in later 2013. This kind of smart contracts are programs in a new “scripting language” that can construct new applications on top of blockchains. The applications can “privatise” the blockchain for various purposes outside the scope of bitcoins. Proof-of-concept implementation of smart contracts for various applications, including financial auditing, sport betting, and music distributions are already available. The latest contender is, indeed, an entire company whose steering is determined with the use of smart contracts for voting on projects to be funded (the so-called DAO). Hence, with smart contracts, agreements can be transferred from the hands of lawyers and paper documents to the digital world using the blockchain as a virtual machine and relying on the distributed trust it enables. However, there is an inherent issue with smart contracts: who can ensure that the smart contracts are correctly programmed? This fundamental question is of paramount importance for legal and financial applications using smart contracts.

In this project, we will extend a program analysis framework for Smart Contracts with Soufflé. The framework will report on rules in the smart contract that are wrongly implemented. Hence, the abstract interpretation framework will prevent the loss of money caused by wrongly implemented rules.

Requirements: programming knowledge in C++ / some basics of logic

Projects supervised by Zhiyong Wang

Multimedia Data Summarization (12 or 18cp)
Multimedia data is becoming the biggest big data as technological advances have made it ever easier to produce multimedia content. For example, hundreds hours of video are uploaded to YouTube every minute. While such wealthy multimedia data is valuable for deriving many insights, it has become extremely time consuming, if not possible, to watch through a large amount of video content. Multimedia data summarization is to produce a concise yet informative version of a given piece of multimedia content, which is highly demanded to assist human beings to discover new knowledge from massive rich multimedia data. This project is to advance this field by developing advanced video content analysis techniques and identifying new applications.

Human Motion Analysis, Modeling, Animation, and Synthesis (12 or 18cp)
People are the focus in most activities; hence investigating human motion has been driven by a wide range of applications such as visual surveillance, 3D animation, novel human computer interaction, sports, and medical diagnosis and treatment. This project is to address a number of challenge issues of this area in realistic scenarios, including human tracking, motion detection, recognition, modeling, animation, and synthesis. Students will gain comprehensive knowledge in computer vision (e.g. object segmentation and tracking, and action/event detection and recognition), 3D modeling, computer graphics, and machine learning.

Predictive Analytics of Big Time Series Data (12 or 18cp)
Big time series data have been collected to derive insights in almost every field, such as the clicking/view behavior of users on social media sites, electricity usage of every household in utility consumption, traffic flow in transportation, to name a few. Being able to predict future state of an event is of great importance for effective planning. For example, social media sites such as YouTube will be able to better distribute popular video content to their caching servers in advance so that users can start watching the videos with minimal delay. This project is to investigate existing algorithms and develop advanced analytic algorithms for higher prediction accuracy.

Remote-sensing Image Analysis (12 or 18cp)
Remote sensing images have played a key role in many fields such as monitoring and protecting our natural environment, improving agriculture, and assessing water quality. Due to the limitation of the current imaging technology, advanced image analysis techniques such as unmixing and classification are needed to better utilize remote sensing images. Meanwhile, the increasing number of massive remote sensing images demands efficient algorithms in order to support timely decision-making. This project is to investigate efficient and effective approaches to address the emerging issues in remote sensing image analysis. Students will develop strong skills in image analysis, machine learning, and data mining.

Audio Data Analysis (12 or 18cp)
Audio data such as speech and music is very important for our daily life. For example, speech data has rich affective information in addition to the spoken content, and music is very powerful to influence the emotion of individuals. Therefore, it is very helpful to understand the full meaning of a given piece of audio data. This project is to investigate intelligent algorithms for such a purpose and explore the value of audio data. Students will develop strong skills in audio data analysis, machine learning, and data mining, while enjoying the beauty of various sounds.

Projects supervised by Chang Xu

Recommendation with diversified interactions between customers and products (18cp)
Most of existing recommendation systems are built upon single interaction between users and products. For example, in MovieLens recommendation system, only ratings that users provide to items are involved in training procedure. In real applications, however, there are multiple different interactions between users and products, such as clicking, viewing, bookmarking, adding into wish lists, purchasing, and finally writing a review for it. To better understand these multiple interactions between users and items will help us to build up more effective recommendation systems. This project aims to develop an effective recommendation system that considers multiple interactions between customers and products from a holistic perspective. The essential methods for recommendation include data processing, mathematical modeling, visualization, and experimental analysis. Both matrix factorization or deep learning related techniques would be investigated in this task.

Preferred Programming Language: Python / Matlab,

Requirements: Strong programming skills and basic knowledge on machine learning/deep learning.

Learning Smaller Deep Neural Networks for Image Classification (18cp)
Nowadays, deep neural networks have enjoyed significant achievements in many computer vision tasks including image classification. However, most of current well-performed networks involve very wide and deep structure and considerable parameters. This induces the demanding storage requirement and the time-consuming inference cost in the low-end computational devices, such as cellphones and tablets. This project aims to shoot this issue by learning a new smaller network which has less parameters. Nevertheless, the training can be difficult due to the small capacity and large depth of the new network. The new network is expected to learn from existing pretrained models for image classification, such as AlexNet and VggNet. The challenge lies in how to use these existing models to help the training on large image classification datasets, such as ImageNet.

Preferred programming language: Matlab/Python.

Requirements: Strong programming skills and basic knowledge on machine learning/deep learning.

Industry projects

To express interest in these industry projects please email your CV to Evelyn Riegler (evelyn.riegler@sydney.edu.au)

Multi User Clinical Holography (MUCH) (12cp)
The University of Sydney Techlab in collaboration with the Westmead Clinical School
The University of Sydney has Begun development of a new Education hub at the Westmead Clinical School and intends to deliver Holographic consultation and PBL systems in this space. Phase 2 of this this project aims to develop a more robust multi user Holographic consultancy interface, leveraging Microsoft HoloLens(s), with customisation to the interface and back end sharing service with a mind to deploy into classrooms. We would also like to extend this to a multi session application where users can choose their room/PBL, and highlight and or annotate objects in the group.

Supervisors:
Jim Cook - Innovation lead, The University of Sydney
David Cook - Academic Lead, Westmead Institute

Specific requirements/skills: Hololens software is primarily developed in Unity and Visual Studio, and therefore C# skills will be of great benefit. Having completed holographic academy would position applicants to succeed. An understanding of the Windows operating system as well as the ability to mock up simple web interfaces and apps will be beneficial. We are not fussy about web backend platforms, and would be happy to supervise MEAN, RAILS, PHP, et all, at the students choice. Human centred thinking is highly valued in the Techlab, and we work within a design thinking framework.

The Trend Report (12cp)
The University of Sydney ICT Techlab
The ICT Techlab at the University of Sydney, has for 4 years been providing knowledge on Emerging trends in the Technology, Political, Social and Legislative space to academics at the University. While Delivering Prototypes in Holography, Virtual reality, Mobile applications and Geolocation solutions, the Techlab has developed an extensive library of tools and demos related to these trends.
The Techlab is ready to deliver an online experience that is very customisable, and allows users to personalise their consumption of the content. Previously this has been laborious and intensive as slide decks or videos were produced on a per use basis. The goal of this project is to develop a web app that is mobile first, but can deliver complex multimedia and information about trends and projects and their relationships to one another.

The database must be able to show trends, projects, partners and their relationships and projects and trends must be able to support multiple pieces of multimedia. The webapp should be able to deliver a story mode based on a group of predefined user journeys. These journeys should be based around the University strategy (education, research, culture) and the digital principles to which they adhere.

Supervisor:
Jim Cook - Innovation lead, The University of Sydney

Specific requirements/skills: A Design thinking approach and skills in front end dev elopement will be a plus. The web application stack can be in any language or frameworks, we have staff who work in Flask, Ruby on Rails, NodeJS and C#, though we encourage unique approaches to complex problems.

Links:
Information about the Project and Techlab:
http://hypecycle.techlab.works
http://projects.techlab.works
https://www.facebook.com/usydtechlab/

A markup language authoring platform for the production of learning resources (12cp)
The University of Sydney ICT Techlab in collaboration with the School of Electrical and Information Engineering
The increasing presence of blended learning environments in which a substantial percentage of resources are online requires agile production of these resources. Conventional approaches rely on models that require laborious processes to first define and then produce the resources. Markup languages streamline the production process by separating the structural definition of content from its post-production, which is usually performed automatically. The ReAuthoring platform is an extension of the Sphinx-doc platform (used by Python to document the language) with enhancements to produce learning resources. The markup language is extended with features to embed multiple choice questions, dual version (instructor, student), videos, and basic submission forms.

Supervisors:
Jim Cook - Innovation lead, The University of Sydney,
Abelardo Pardo - Academic Lead, School of Electrical and Information Engineering, The University of Sydney.

Specific requirements/skills: The ReAuthoring and sphinx doc platform is written in Python. Therefore, strong programming skills in Python and understanding of compilation processes are required. The main contribution of the project is to improve the post production process to obtain a high-quality user experience. For this purpose, experience with toolkits for UI design such as bootstrap or similar will be required. The ability to extend and iterate an already existing product in quick cycles will also be beneficial. The enhancements derived from the project can be tested in live environments with real users, so the quality of the final product has to be very high.

Reference Material/Links (if applicable):
ReAuthoring
Sphinx-doc
Sphinx Themes

Development of a workflow program to validate compounding techniques to make medicines using the HoloLens® (12cp)
The University of Sydney ICT Techlab in collaboration with the School of Life and Environmental Sciences
Compounding medicines in pharmacy can range from simple creams to high-risk sterile products to be injected directly into patients. Errors in the compounding process can have serious consequences, resulting in patient injury and, in some cases, death. It is hypothesized that a high-quality compounding workflow and quality checks will aid prevent compounding errors from reaching patients. We propose to pilot test the use of the HoloLens® to assist in compounding medicines by pharmacists and pharmacy students. The HoloLens® will display the standard procedure for compounding and any relevant clinical information. In addition, it will include voice-activated image capture system to match against each step on the standard procedure for improved documentation and off-site verification. A barcode or image verification system may also be developed.

Strategy for developing a workflow program for compounding medicines
Create an Augmented Reality Application: Augmented reality hardware, such as the HoloLens® has the potential to improve visual documentation while compounding medicines without interrupting workflow. The function of the software is to display specified standard procedures and relevant clinical information. An auditing step is to capture an image to align with each step of the standard procedure. Additional verification of the image or barcode to match the ingredients listed in the compound is recommended.

Focus on education: Current pharmacy students require experience and practice in compounding medicines. Using the augmented reality software students can practice and train their skills. Furthermore, students will be able to identify areas of weakness which they can they focus on to improve. Finally, the auditing aspect of the software and technology can be expanded to student assessments.

Supervisors:
Jim Cook - Innovation lead, The University of Sydney
Jonathan Penm - Academic Lead, Faculty of Pharmacy
Ardalan Mirzaei - Industry supervisor, Faculty of Pharmacy.

Development of a workflow program for real-time in-clinic visualization of paediatric cardiac MRIs in Oculus Rift (12cp)
The University of Sydney ICT Techlab in collaboration with the School of Medical Sciences and the Children’s Hospital Westmead
MRI scans for cardiac patients are routine for identifying defects in the heart and for determining any surgical interventions to correct the defect. This can be more challenging in the case of paediatric patients where the heart is relatively small. In some more difficult cases, the clinicians were not able to find the defect in the MRI scans, and as a result had to resort to 3D printing the heart which, although allowing successful identification of the defect, resulted in a +12 hour wait while the print was completed. We recently trialled an Oculus/Unity based approach which allowed us to import the 3D data and allow the clinician to complete a detailed investigation of the heart within 10 min in his rooms. The purpose of this project is to refine the MRI visualization and transfer to Oculus that can be incorporated into the normal workflow of the clinician. This will also allow the clinician to provide detailed explanations to the patients and family as to the nature of the defect and proposed procedure - via the use of shared Oculus visualization and to highlight areas of the tissue using Tilt Brush. In addition to its role in diagnosis, there is a very strong enthusiasm for developing Oculus as a powerful educational tool to teach the anatomy of the cardiovascular system.

Strategy for developing a workflow program for VR cardiac imaging
Create an Virtual Reality Workflow: The main aim of the project is to work together with the clinician, biomed engineers and medical science student to develop the work flow that will allow a clinician sitting at their desk to import the MRI scan data into Oculus Rift and to navigate their way around the heart using Touch. Some programming is required to adapt the scanning software Further refinements include the use of Tilt Brush and exploring networking multiple headsets or other opportunities for multi-user interaction.

Focus on education: This project addresses the educational needs of two cohorts. The first is the patient and family. Through the use of this technology they will be provided with an unparalleled insight into the defect and how it affects heart function and how it can be treated with surgery. The second is the medical professionals, from the intern to the senior specialist, this technology provides a new and exciting platform to observe, learn and think about the cardiovascular in new ways that can also be used in undergraduate teaching. This in itself is an important research project.

Supervisors:
Jim Cook - Innovation lead, The University of Sydney
Philip Poronnik - Professor of Biomedical Sciences, School of Medical Sciences
David Winlaw - Professor of Peadiatric cardiology, Children’s Hospital Westmead
Tegan Cheng - Lecturer in Biomedical Engineering, EPIC Centre Westmead

Medication and Prescription Cross-Checking Optimisation (12cp)
The University of Sydney ICT Techlab in Collaboration with the Faculty of Pharmacy
Doctors write prescriptions for patients which can then be dispensed at a pharmacy. Minor differences in dose and strength can cause serious harm to patients and even death. As part of the work flow of dispensing a prescription, pharmacists will perform a final cross-check with the medication dispensed and the doctors’ original prescription.

In order to balance cross-checking speed, accuracy and precision with clinical and medication suitability, technological advancements can be utilised for this process. We desire developing software that can function with augmented reality technology. This software can assist pharmacists by automatically cross-checking the prescription and medication dispensed, allowing pharmacists to dispense with increased accuracy and focus on other clinical assessments.

Strategy for using technology to reduce cognitive burden on pharmacist and improve cross-checking accuracy.
Create an Augmented Reality Application: Augmented reality hardware, such as the HoloLens® has the potential to facilitate in the cross-checking procedure. The function of the software is to read the prescription, read the dispensed label, inspect the medication, and cross-check that all three corroborate with each other. A final auditing step is to capture an image. Current auditing relies on the medication barcode and the dispensing software, whereas, with image capture, batch numbers and medication expiry’s can also be stored.
Focus on education: Current pharmacy students require experience and practice in cross-checking procedures. Using the augmented reality software students can practice and train their skills. Furthermore, different aspects of the cross-checking procedure can be adjusted for practice purposes. This is so as not to encourage complete reliance on technology, but rather to hone in on cross-checking skills that may need be further developed in some students. Finally, the auditing aspect of the software and technology can be expanded to student assessments.

Supervisors:
Jim Cook - Innovation lead, The University of Sydney,
Jonathan Penm - Academic lead, Faculty of Pharmacy,
Ardalan Mirzaei - Industry supervisor, Faculty of Pharmacy

Sharing Haswell’s educational legacy and the University’s scientific heritage (12cp)
The University of Sydney ICT Techlab in collaboration with the School of Life and Environmental Sciences
The Haswell zoological collection (museum) was established in 1903 by Prof William Haswell, the first Challis Professor of Zoology at the University of Sydney. For over 110 years this collection, consisting of invertebrate and vertebrate specimens, has supported teaching and research in biology.

We have recently completed an audit of the Haswell museum and created the ‘pilot’ Haswell database in FileMakerPro system, which has allowed for an efficient auditing process (audit = photograph and create database entries). This database has already supported colleagues within the School of Life Sciences and Macleay Museum who have used the pilot database as a source of inspiration for upcoming classes, outreach activities and research exhibitions.

Strategy for sharing Haswell’s educational legacy and the University’s scientific heritage
Create a sustainable database solution: FileMakerPro, whilst useful for the audit, does not offer a long-term solution for sharing the Haswell collection outside of the University. To ensure maximum access to the collection for research and teaching we require a database aligned to zoological taxonomy (replicating the FileMakerPro database) that includesa role-based object booking system, and a user-friendly web ‘front-end’ (integrating the booking system and user searches). In a project with Computer Science students in the second half of 2016 we made reasonable progress on creating the booking system, this system needs to be refined, integrated with a non-filemakepro database, and a frontend designed.

Focus on education: By having the records in the Haswell collection digitised as 3D digital ‘objects’, the potential research and educational user-base can be extended. We have linked the pilot FilemakerPro database to the Intermediate Zoology Online resources. 3D technologies will enhance the value of the collection to students and visitors; 3D scanning options include laser scanning (resolution ~ 0.1 mm), photogrammetry (resolution ~0.1 mm), X-ray/CT/microCT (various technologies, resolution ~ 0.01 mm). Commenced; exemplars of CT scans and .gifs are being rendered.

Supervisors:
Jim Cook - Innovation lead, The University of Sydney
Rosanne Quinnell - Academic Lead, School Life and Environmental Sciences

Specific requirements/skills: Database creation to mimic the phylogenetic structure of the existing FileMakerPro database. This database application must be user-friendly for biology curators. The database information shall be presented onto a website accessible by the public. Human-centred thinking is highly valued and we work within a design thinking framework.

Links:
Information about the Haswell collection:
SOLES webpage
Worldwide Database of
University Museums and Collections

Atlas of Living Australia
Pilot Haswell database in FileMakerPro - please note to access you must have a unikey login and be on the University intranet (or VPN).
Digitisation Project website
Twitter
Instagram

Quality Assurance System Using Machine Learning & Computer Vision (12cp)
WipeHero
Here at WipeHero, we’ve seen our number of washes go from less than ten to many thousands per month, all in the span of half a year. With this number continuing to rapidly rise, the need for an automated system to assist with the auditing and quality assurance of jobs has arisen.

The proposed solution is a machine-learning / computer vision based system which reviews photos submitted by WipeHero’s car washers. The system will be able to recognise both interior and exterior sections of vehicles, such as seats, cup holders, dashboards, and exterior surfaces.

Knowledge/experience with any of the following would be very beneficial: Assisted machine learning / image recognition MATLAB. Work with any popular ML/image libraries such as Pytorch, Tensorflow, or OpenCV

Benefits of working with us include:

  • Become involved in Australia's fastest growing on-demand carwash service
  • Open, respectful, fun company culture
  • Strong focus on your professional development, possible further employment opportunities beyond the semester program
  • Team lunches and free carwashes from our lovely washing heroes

About WipeHero
Founded by Harvard and USyd/UNSW graduates, WipeHero is a Sydney-based startup focused on making it easier for people to get services for their cars, starting with car cleaning! We’ve developed a waterless technology that enables our washing heroes to wash cars anywhere without a single drop of water, saving customers time and the environment up to 200 litres of water per wash. WipeHero has rapidly emerged as the leading service provider for individuals and fleets, with backing from top tier investors.

Identification of common cluster of diseases associated with primary diagnosis of a heart condition and analysis of variations in costs and length of stay for patients with different clusters of diseases (12cp)
CMCRC – Health Market Quality Program
The Health Market Quality (HMQ) R&D program of the Capital Markets Cooperative Research Centre (CMCRC) covers the application of advanced data sciences to all public and private healthcare settings with the objective of improving integrity and efficiency for all stakeholders in the market.

The student will be working with anonymized health data provided by one of CMCRCs industry partners and is expected to work with data stored in a secure environment within CMCRC.

The project involves analysing large volumes of de-identified hospital data to study the variations in treatment, costs and length of stay for the different clusters of patients. This requires
(a) understanding what secondary diagnoses are most commonly clustered for patients with primary diagnosis of an acute condition (such Acute Myocardial infarction).
(b) clustering patient groups based on diagnosis types to identify multiple clusters of patient types.

The student is expected to develop the analytical process to deliver results that can be depicted through informative graphs and charts that can be interpreted by hospital managers who are not trained in statistics.

The results of the analysis should be written up succinctly for presentation to hospital managers and health industry stakeholders.
In-house training will be provided to help the student understand the data and the different health information codes present in the dataset.

Supervisors:
Dr Uma Srinivasan - CMCRC – Health Market Quality Program
Associate Professor Bernhard Scholz - School of IT, The University of Sydney

Desirable skills : exposure to the health domain

Specific requirements/skills: The student should be familiar with statistical or machine learning techniques and be proficient in a statistical programming language such as R, Python or Strata.

The student should have a good command of spoken and written English and be able to write up the analysis findings in a report format that can be presented to stake holders in the health system.

Reference material/links (if applicable):
https://www.cmcrc.com/health

Timeframe: Duration of project: Monday 31 July 2017 – Friday 24 November 2017

Important Note: Please note the results of the analysis cannot be published unless the student and the supervisors have ethics clearance from the University.

Study of variation in length of stay and cost of care over time, for high cost surgical procedures for patients with different health profiles (12cp)
CMCRC – Health Market Quality Program
The Health Market Quality (HMQ) R&D program of the Capital Markets Cooperative Research Centre (CMCRC) covers the application of advanced data sciences to all public and private healthcare settings with the objective of improving integrity and efficiency for all stakeholders in the market.

The student will be working with anonymized health data provided by one of CMCRCs industry partners and is expected to work with data stored in a secure environment within CMCRC.

The project involves analyzing large volumes of de-identified hospital data to understand the variation in cost of care over time, for different high cost surgical procedures (such as hip replacement) for patients with varying health profiles indicated by:

(a) the nature of their presentation at the hospital (emergency vs planned admission),
(b) their primary and secondary diagnosis codes. (e.g., patients with chronic conditions such as diabetes).

The student is expected to develop the analytical process to deliver results that can be depicted through informative graphs and charts that can be interpreted by hospital managers who are not trained in statistics.

The results of the analysis should be written up succinctly for presentation to hospital managers and health industry stakeholders.
In-house training will be provided to help the student understand the data and the different health information codes present in the dataset.

Supervisors:
Dr Uma Srinivasan - CMCRC – Health Market Quality Program
Associate Professor Bernhard Scholz - School of IT, The University of Sydney

Desirable skills: exposure to the health domain

Specific requirements/skills: The student should be familiar with statistical or machine learning techniques and be proficient in a statistical programming language such as R, Python or Strata.

The student should have a good command of spoken and written English and be able to write up the analysis findings in a report format that can be presented to stake holders in the health system.

Reference material/links:
https://www.cmcrc.com/health

Timeframe: Duration of project: Monday 31 July 2017 – Friday 24 November 2017

Important Note: Please note the results of the analysis cannot be published unless the student and the supervisors have ethics clearance from the University.

Online Authorized Entry Signage generator ALSO populates Hazardous Areas register (12cp)
The University of Sydney
Develop a web based application that generates the University’s Authorized Entry only signage AND integrates the hazard data collected with an existing university information system, to help inform and roll out work health and safety strategies, advise building and maintenance personnel and control access.

Background
Access to areas posing risks to health and safety, such as laboratories and workshops, have restricted access. Only staff, students and contractors who have been inducted to these areas are permitted to enter unsupervised. Everyone else must be under direct supervision.

These restricted areas are identified by standardised Authorised Entry Only signage. This signage is implemented where a risk of harm exits or a state of warning or regulatory compliance is required.

This signage displays basic information on:

1. Hazards in the area
2. Safety precautions to be taken when working in the area
3. Contact details for the area supervisor
4. Contact details of the local emergency warden
5. Contact details of the local first aid officer

Current process
Area supervisors use a Microsoft Word template with some Guidelines, to print out on A3 paper and display at entrances to hazardous areas.

Proposed enhancements
Area supervisors use an online tool to select and enter the basic information for the signage (see above numbered items). This information is used to layout the required pictograms and contact information into a sign in PDF format AND create (or modify) a record for that area within the University’s building management system – Archibus.

Stakeholders
School of Information technology – the school within the Faculty of Engineering and Information Technology, assessing project proposals and linking them to students to provide IT solutions.

Safety Health & Wellbeing (SHW) – the centralised university department responsible for the university’s safety management system

Campus Infrastructure Services (CIS) – the centralised university department responsible for the built environment and the University’s building management system; Archibus.

Information Communications Technology (ICT) – the centralised university department responsible for the university’s IT infrastructure.

Our Pills Talk Medication Safety App (12cp)
Our Pills Talk Pty Ltd
"Our Pills Talk Medication Safety App" project, for both iOS and Android devices, requires various embellishments to the app, the website, and the visual presentation:

These will include as below, if able, but may include other options, should these ensue during our program:

  • Embellishment progress of the existing app as per (a) us, the clients, and (b) student innovative concepts, ideas and embellishments
  • Improving quality of multiple language translations of text to speech, particularly relating to medication descriptions eg capsule, tablet, cream, inhaler, eye drops etc. This allows translating of the doctor’s instructions on the patient’s medication, into their preferred language. Ideally using Microsoft Azure software, to maximise the number of languages available for translating
  • To allow the dispensed drugs to be automatically texted (preferably at no cost to either the dispensing pharmacist or the patient), or by a free cloud concept, or by any other available option, with the patient’s permission, to appear in the patient’s scanned HISTORY Tab,
  • Alert reminder for patient to take their medication
  • Alert reminder for the patient to have their prescription repeat dispensed
  • Improved links of the patient’s drugs to their CMI (Consumer Medicine Information as on the NPS.gov.au website
  • Add links to external Health apps eg. government website: HealthDirect.org.au
  • Allowing collection of scanned data using eg MS SQL Serve, ideally MS SQL Express, or, otherwise MySQL
  • Create a PowerPoint presentation of How To Use "Our Pills Talk Medication Safety App"
  • Create a demo video for Android and iOS.
  • possible upgrading of OurPillsTalk.com.au website
  • Possible option: Assist with applying for funding from eg Angel Investors Crowd funding Government Grants etc
  • Possible option: Business Plan to forward to these potential investors

Supervisors:
Steve Cohen, pharmacist, Supervisor
Michelle Spiro, pharmacist colleague, as Collaborating Supervisor

Specific requirements/skills:

  • App embellishment skills with iOS and Android platforms on phones, and also ideally on tablet devices
  • PowerPoint creations
  • IT skills with MicroSoft Azure text to speech language translations
  • collecting data and statistics of scanned medications
  • basic website embellishment skills
  • embellishment of, and creating Demo videos for, OurPillsTalk.com.au website