Undergraduate Projects - 2012
In 2012 I will be offering projects on three topics:
1) Text Mining, 2) Web Applications 3) Affective computing 4) Gestural games (e.g. XBox Kinect). If you are interested in any of these topics, have good grades or other qualifications send me a line.
Some of these projects are industry based, this means that most of the supervision might fall in the industry partner, and in some cases it might require that you work on their premises.
RAC1. Evaluation of classification algorithms for Affective Computing
Affective computing aims to make computers that recognize emotions from physiological signals, text, facial expressions etc. In this project you will develop machine learning classifiers that make sense of this data and can be used to detect emotions.
Read: R.A. Calvo and S. D'Mello (2010). "Affect Detection: An Interdisciplinary Review of Models,Methods, and their Applications" IEEE Transactions on Affective Computing. 1(1), 18-37.
RAC2. Facial expression recognition with the XBox Kinect
The Kinect camera has been one of the fastest selling digital devices.
In this prject you will study how facial expressions can be detected from video recorded from a Kinect camera. You will explore its applications in video games.
Read: 3D corpus of spontaneous complex mental states
M. Mahmoud, T. BaltruĊĦaitis, P. Robinson, L. Riek (2011)
Affective Computing and Intelligent Interaction (ACII)
RAC3. Gesture recognition with the XBox Kinect
The Kinect camera has been one of the fastest selling digital devices.
In this prject you will study how gestures can be detected from video recorded from a Kinect camera. You will explore its applications in video games.
Read: 3D corpus of spontaneous complex mental states
M. Mahmoud, T. BaltruĊĦaitis, P. Robinson, L. Riek (2011).
Affective Computing and Intelligent Interaction (ACII)
RAC7. Electrophysiological signals for the recognition of emotion
When you fill an emotion your body changes (e.g. if you are scared your heart might accelerate). You will collect and analyse electrocardiogram, electromyogram and skin conductivity recordings to classify emotions. You will evaluate a number of classification algorithms. Experience with signal processing or data mining will be a plus.
RAC 8: Emotional expressions and improvement of Race Car simulators
In this project you will work with a race car simulator to create situations (e.g. an accident) that trigger affective responses: surprise, engagement...
Read:
Affective Videogames: The Problem of Wearability and Comfort. A. Bonarini, F. Costa, M. Garbarino, M. Matteucci, M. Romero, S. Tognetti (2011). Human-Computer Interaction. Users and Applications p. 649–658
RAC10: Web 2.0 Campaigning for non-profits
Good Return is a website that engages and connects the Australian public with low income borrowers in Australia and Asia Pacific that require microfinance loans to lift themselves out of poverty. The Good Return initiative has been developed by World Education Australia, an international not for profit that focuses on literacy, livelihood and microfinance. The Good Return website is currently in beta mode and as we prepare for public launch we require assistance in the implementation of an SEO plan. Your role will be to implement the SEO plan and write a facebook, twitter application and others. It is particularly recommended when you can take Elec5619 concurrently.
RAC 11. Industry Projects with Freelancer.com
Freelancer offers a $1000 reward to good applicants with at least a credit average. You may also want to discuss a paid internship with Freelancer. These arrangements are between you and the company, the university is not involved in any way. Please email your CV to .
Realtime Fraud Detection using Machine Learning:
This project will use machine learning techniques to detect fraudulent users and fraud activity on an extremely large userbase with thousands of transactions per second. Online fraud detection is one of the largest security problems on the Internet, costing organisations billions of dollars annually.
This project will involve data mining a large website with complex and constantly evolving fraud and developing a system that detects it and the relationships between fraudulent parties.
Realtime Advertising Optimisation using Machine Learning
This project will use machine learning techniques to optimise realtime advertising campaigns using Google Adwords. The system will be designed to develop optimal advertising copy and keywords
from datamining, then train with realtime feedback on performance provided by analytics and billing information. This will be a real world ad campaign and not a simulation. You will know exactly how much revenue you generated and the ROI of your campaigns at the end of the thesis.
Simply put, the system will be a black box for making money.
Automated Front-end Testing Framework for Web Applications:
With the Internet evolving, websites are so complex that manually testing each and every form of user input is impossible. Your project will be to create an automated framework for black box testing large websites with complex input scenarios. The system will be modular and scalable.
As an extension, the system should be able to detect when a page changes and automatically
evolve the test.
Design of High Performance Search for a Massive Multi-content Website:
Search is one of the great problems of today which has resulting in Google becoming a multibillion dollar company, and dominating the Internet. A parallel problem is how to search existing databases, which have not been designed with search included in the initial design. This project
will be to develop a search system, harnessing existing technologies and tools to provide high performance search across massive databases holding a variety of content.
# Recommender systems
A recommender system compares user profiles to some reference characteristics, and seeks to predict the 'preference' that user would give to an item they had not yet considered. We are interested in exploring both content-based and collaborative filtering approaches for recommendation for the purpose of both employer and freelancer project recommendations.
# Visualisation
We have massive amounts of data and are in need of scalable visualisation tools. We aren't just looking for boring financial charts, work with us to research and develop interactive visualisations of live user events (projects posted, completed, new registrations, logins, etc). We are also interested in developing ways to visualise our social network of over 2.5 million vertices.
# Data Mining
We are investigating the use of data mining techniques to discover interesting patterns within our massive datasets. We are currently investigating the use of classification techniques for fraud analysis and are interested in applying graph mining techniques for a similar purpose.
# Web Development
Work with the best in the business to develop a new idea from start to finish. Be responsible for a feature that will be used by over 2.5 million users worldwide! Strong experience in web programming and SQL are a must.
RAC8. Recognizing emotions in text.
When you write an emotionally charged document the vocabulary you use might change. Think about the way you write in Facebook or your emails, how is it different from the way your write an assignment? Could you detect automatcially what is the mood of someone from what they write on Facebook? Twitter?
In this project you will explore approaches to classifying emotions in text.
Read:
- R. A. Calvo and M. Kim (to appear) "Emotions in text: dimensional and categorical models". Computational Intelligence
- Sunghwan Mac Kim, A. Valituti, R.A. Calvo "Evaluation of Unsupervised Emotion Models to Textual Affect Recognition". Workshop in Computational Approaches to Analysis and Generation of Emotion in Text, part of the Human Language Technologies: The 11th Annual Conference of the North American Chapter of the Association for Computational Linguistics, June 1–6, 2010, Los Angeles
RAC 9. GeoAffective maps
Build maps that describe what people feel in different locations. This would be a combination of Google Maps with Flickr and an emotional annotation protocol. It is particularly recommended when you can take Elec5619 concurrently.
iFarm: A virtual farm for training veterinary and animal science students in design
Designing field experiments or clinical trials is an essential skill for animal and veterinary science graduates but it is difficult to impart this skill due to logistic reasons. This project will develop a virtual farm program to enable students design and conduct field/clinical trials mimicking real life scenarios. This program is also aimed to be a fun activity for students to improve their engagement in learning activities.
Your role will be to design a fully functional program and a concise program manual for the students. You will develop the program for one or two field trials for one animal species but the program should be flexible to enable design of trials for other animal species in the future. Dr Navneet Dhand and Associate Professor Peter Thomson of the Faculty of Veterinary Science will act as your clients and you will be supervised by Associate Professor Rafael Calvo.
The student involved in the project will be eligible to receive funding to present the developed program at a conference/meeting in Australia or overseas (maximum limit of $3000). This money can also be used for other academic purposes with the clients’ consent.