Big Media Intelligence Super Project - Sem 2, 2020
This super project consists of a series of independent tasks in large scale multimedia data analytics. Students will select ONE task from any research directions to undertake their thesis project.
Research Direction 1 - Ambulance Data Analysis and Visualization
Task 1 - Ambulance Data Analysis and Visualization
Task 1.1: Data visualization: In this task, students are required to visualize the compelling stories by transforming the data and decide the right type of visualization. Besides, a dashboard is expected to be designed for demonstrating this dataset on a user-friendly interface. Students participating in this project are expected to have the following skills:
- good Web development skills including both frontend and backend system development. A dashboard is expected to be implemented.
- good understanding of database management
- experiences of data mining (INFS4203/7203) or related areas
Task 1.2: Data Analysis: Before conducting data analysis, detecting and correcting corrupt or inaccurate records has to be done on the dataset, i.e., data cleaning. With the goal of discovering useful information, informing conclusions and supporting decision-making, students are required to have in-depth understanding of the interesting patterns inside the provided dataset.
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), or pattern recognition and analysis (INFS3710)
- good programming skills
Task 2 - Traffic Data Analysis and Visualization
Task 2.1: Data Visualization. In this task, the students are expected to (1) visualize the real world traffic data (i.e., goCard transactions) provided by TransLink and help staff to understand the passenger flow; (2) Find the relevance between traffic data and social events (e.g., major uni events). Students participating in this project are expected to have the following skills:
- good Web development skills including both frontend and backend system development. A dashboard is expected to be implemented
- good understanding of database management
- experiences of data mining (INFS4203/7203) or related areas
Task 2.2: Traffic Prediction. Generally, students are required to model session-based traffic data with sequential model or graph model and estimate the variance of people flow. Students participating in this project are expected to have the following skills:
- experiences of machine learning (COMP4702), data mining (INFS4203/7203; INFS7450), high-dimensional data processing (INFS4205/7205), information retrieval (INFS7410) or pattern recognition and analysis (INFS3710)
- strong programming skills
Research Direction 2 - Content-based Image Understanding
Task 3 - Curb & Channel Crack Detection
Task 3.1: Road Scene Segmentation: Students are required to study pixel-labeling approaches using semantic segmentation for road scene understanding. Students participating in this project are expected to have the following skills:
- image processing
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), or pattern recognition and analysis (INFS3710)
- Strong programming skills
Task 3.2: Crack Detection: Automatic detection of pavement cracks is an important task in transportation maintenance for driving safety assurance.However, it remains a challenging task due to the intensity inhomogeneity of cracks and complexity of the background, e.g., the low contrast with surrounding pavement and possible shadows with similar intensity. Students participating in this project are expected to have the following skills:
- image processing
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), pattern recognition and analysis (INFS3710)
- Strong programming skills
Research Direction 3 - Social Data Analysis and Visualization
Task 4 - Event Detection
Crawl and analyse news or tweets for public event detection. A dashboard (or other interactive methods) is expected to be designed for demonstrating the event.
- Web development skills including both frontend and backend system development. A dashboard is expected to be implemented.
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205)
Task 5 - Sentiment Detection
Crawl and analyse news or tweets for event sentiment analysis. This will be focused on the the analysis of public reactions to big social events. Students are required to visualize the compelling stories by transforming the data and decide the right type of visualization. Students participating in this project are expected to have the following skills:
- Web development skills including both frontend and backend system development. A dashboard is expected to be implemented.
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205)
Task 6 - Community Profiling
Crawl and analyse news or tweets for community analysis. This will be focused on graph construction, summarisation and profiling. Students are required to visualize the compelling stories by transforming the data and decide the right type of visualization. Students participating in this project are expected to have the following skills:
- Web development skills including both frontend and backend system development. A dashboard is expected to be implemented.
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205)
Big Media Intelligence Super Project - Sem 1, 2020
This super project consists of a series of independent tasks in large scale multimedia data analytics. Students will select ONE task from any research directions to undertake their thesis project.
Research Direction 1 - Ambulance Data Analysis and Visualization
Task 1 - Ambulance Data Analysis and Visualization
Task 1.1: Data visualization: In this task, students are required to visualize the compelling stories by transforming the data and decide the right type of visualization. Besides, a dashboard is expected to be designed for demonstrating this dataset on a user-friendly interface. Students participating in this project are expected to have the following skills:
- good Web development skills including both frontend and backend system development. A dashboard is expected to be implemented.
- good understanding of database management
- experiences of data mining (INFS4203/7203) or related areas
Task 1.2: Data Analysis: Before conducting data analysis, detecting and correcting corrupt or inaccurate records has to be done on the dataset, i.e., data cleaning. With the goal of discovering useful information, informing conclusions and supporting decision-making, students are required to have in-depth understanding of the interesting patterns inside the provided dataset.
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), or pattern recognition and analysis (INFS3710)
- good programming skills
Task 2 - Traffic Data Analysis and Visualization
Task 2.1: Data Visualization. In this task, the students are expected to (1) visualize the real world traffic data (i.e., goCard transactions) provided by TransLink and help staff to understand the passenger flow; (2) Find the relevance between traffic data and social events (e.g., major uni events). Students participating in this project are expected to have the following skills:
- good Web development skills including both frontend and backend system development. A dashboard is expected to be implemented
- good understanding of database management
- experiences of data mining (INFS4203/7203) or related areas
Task 2.2: Traffic Prediction. Generally, students are required to model session-based traffic data with sequential model or graph model and estimate the variance of people flow. Students participating in this project are expected to have the following skills:
- experiences of machine learning (COMP4702), data mining (INFS4203/7203; INFS7450), high-dimensional data processing (INFS4205/7205), information retrieval (INFS7410) or pattern recognition and analysis (INFS3710)
- strong programming skills
Research Direction 2 - Content-based Image Understanding
Task 3 - Curb & Channel Crack Detection
Task 3.1: Road Scene Segmentation: Students are required to study pixel-labeling approaches using semantic segmentation for road scene understanding. Students participating in this project are expected to have the following skills:
- image processing
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), or pattern recognition and analysis (INFS3710)
- Strong programming skills
Task 3.2: Crack Detection: Automatic detection of pavement cracks is an important task in transportation maintenance for driving safety assurance.However, it remains a challenging task due to the intensity inhomogeneity of cracks and complexity of the background, e.g., the low contrast with surrounding pavement and possible shadows with similar intensity. Students participating in this project are expected to have the following skills:
- image processing
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), pattern recognition and analysis (INFS3710)
- Strong programming skills
Research Direction 3 - Deep Learning for Video Generation
Task 4 - Everybody Dance Now
Task 4.1 Dancing Video Generation Given a source picture (video) of a person dancing, this project aims to transfer that performance to a novel (amateur) target performing the same moves. Students participating in this project will investigate applying the Generative Adversarial Nets (GAN) to synthesize the target video based on the pose estimation from the source video and the feature extraction from the target image.
- Deep Learning
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), pattern recognition and analysis (INFS3710)
- https://arxiv.org/abs/1808.07371, https://github.com/GordonRen/pose2pose
Task 5 - Super Slomo
Task 5.1 SlowMotion Video Generation Given two consecutive frames, video interpolation aims at generating intermediate frame(s) to form both spatially and temporally coherent video sequences. The purpose is to interpolate natural images between consecutive frames in a normal video. Students participating in this project will investigate applying Generative Adversarial Nets (GAN) to generate reasonable interval frames.
- Deep Learning
- experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), pattern recognition and analysis (INFS3710)
- https://arxiv.org/abs/1712.00080, https://github.com/avinashpaliwal/Super-SloMo
Big Media Intelligence Super Project - Sem 2, 2019
15 students maxThis super project consists of a series of independent tasks in large scale multimedia data analytics. Students will select ONE task from any research directions to undertake their thesis project.
Research Direction 1 - 3D Data Generation and Analysis
Task 1.1: Dataset Construction: In this task, the students are expected to (1) learn how to use Kinect and/or 3D scanner; (2) collect 3D meta-data for target objects (3) clean the data and store data into database (4) analyse statistics of collected datasets compared with existing datasets. Students participating in this project are expected to have the following skills: frontend (or) backend system development experiences, or experiences in 3D data processing, or experiences in image processing, and good programming skills
Task 1.2: Pattern Recognition for 3D Data: Generally, students are required to understand and improve state-of-the-art 3D object retrieval / 3D reconstruction algorithms. Students participating in this project are expected to have the following skills: experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205) strong programming skills
Research Direction 2 - Traffic Data Analysis
Task 2.1: Data Visualization: In this task, the students are expected to (1) visualize the real world traffic data provided by TransLink and help staffs to understand people flow; (2) Find the relevance between traffic data and social events (e.g., major uni events). Students participating in this project are expected to have the following skills: frontend (or) backend system development experiences (for data visualisation. A dashboard is expected to be implemented) good understanding of database management data mining (INFS4203/7203) good programming skills
Task 2.2: Traffic Prediction: Generally, students are required to model session-based traffic data with sequential model or graph model and estimate the variance of people flow. Students participating in this project are expected to have the following skills: experiences of machine learning (COMP4702), data mining (INFS4203/7203; INFS7450), high-dimensional data processing (INFS4205/7205), information retrieval (INFS7410) pattern recognition and analysis (INFS3710) strong programming skills
Research Direction 3 - E-health Data Analysis
Task 3.1: Sleep Data Analysis: Predict the sleep status (e.g., non-rapid eye movement (NREM)) with provided sensor data. Students participating in this project are expected to have the following skills: experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), pattern recognition and analysis (INFS3710) strong programming skills
Task 3.2: EGG Data Analysis: The students are expected to use electroencephalogram (EEG) data to detect fatigue during driving or under other scenarios. Students participating in this project are expected to have the following skills: experiences of machine learning (COMP4702), data mining (INFS4203/7203), high-dimensional data processing (INFS4205/7205), pattern recognition and analysis (INFS3710) strong programming skills
Research Direction 4 - Automatic Scientific Literature Analysis
Task 4.1: The students are expected to develop an intelligent scientific data exploration software that leverages natural language processing, data mining and machine learning techniques for automatic scientific literature analysis and implicit knowledge discovery. Students participating in this project are expected to have the following skills: frontend (or) backend system development experiences (for data visualisation. A dashboard is expected to be implemented) good understanding of database management data mining (INFS4203/7203; INFS7450) information retrieval (INFS7410) good programming skills
Hayley Faulkner
Now Pose! Makeup Transformation
This thesis explores various methods of improving the performance of a CycleGAN makeup transfer network on faces with pose. Generally, face transformation networks are mostly trained on fully-frontal images and lack the exibility to be able to perform well on faces with pose or profile images. Speciffically, this project investigates the impact of training with combinations of datasets with varying pose diversity, the affect of the generator architecture and the addition of facial feature segmentation masks to aid the network in it's transformations.
Learn MoreJunhao Lin
Online Action Detection and Multi-Task Predictions with Kinect and Deep Learning
This work investigates online action analysis which is a crucial component for various real-time applications, based on the Kinect sensor and the popular deep learning approach these days. Skeleton data is the chosen modality rather than RGB or depth data, because of its low dimensionality as well as effectiveness shown in previous researches. In terms of neural networks, a RNN-based model has been selected which is suitable for sequence learning while having probably less complexity and shorter latency than CNN-based models.
Nhan Tri Luong
Action Recognition by Pose Estimation
His study investigates the sibility of using machine learning models to automate these processes. It works by training the models to recognize human actions in the videos and useiJ output to categorize videos or block offensive videos that contain inappropriate actions. The results showed that by combining different architectures, the accuracy of the whole system increased because more information was extracted from the videos.
Zijian Wang
Deep Collaborative Supervised Hashing for Efficient Image Retrieval
This work investigates online action analysis which is a crucial component for various real-time applications, based on the Kinect sensor and the popular deep learning approach these days. Skeleton data is the chosen modality rather than RGB or depth data, because of its low dimensionality as well as effectiveness shown in previous researches.
Ryan Garthwaite
Temporal Action Recognition Via Pose Estimation
Action recognition from videos is an active area of research tackling wide ranging problems from violence in CCTV streams to monitoring disabled or elderly people for urgent medical situations. The goal of this project is to apply techniques of action recognition to the situation of human gesture recognition commonly performed in public speaking situations. The actions of clap, cough, wave and point are detected within untrimmed videos that may include multiple instances of these actions.
Shunjia Tao
An Intelligent Fashion Closet System
This study develops e-commercial website called fashion closet. Aside from developed user interface with general online shopping features, the recommendation algorithm is also developed for assisting people to get recommendation and find the clothes that close to their preference more easily.
Team members
Mahsa Baktashmotlagh
Senior Lecturer
Xin Yu
Senior Lecturer
Sen Wang
Senior Lecturer
Yadan Luo
Lecturer
Ruihong Qiu
Zhi Chen
Pengfei Zhang
Zijian Wang
Current Research Students
Jason Zhao
Boyu Luo
Bowen Yuan
Danny Wang
Yan Jiang
Yanran Tang
Tianqi Wei
Yilun Allen Liu
Yiyun Zhang
Ivan (Zhuoxiao) Chen
Cross-disciplinary Collaborators
Professor Scott Chapman
School of Agriculture and Food Sustainability
Faculty of Science
University of Queensland
Professor Han Huang
Faculty of Engineering, Architecture and Information Technology
University of Queensland
Professor Lianzhou Wang
Australian Institute for Bioengineering and Nanotechnology
Faculty of Engineering, Architecture and Information Technology
University of Queensland
Professor John Zhu
Faculty of Engineering, Architecture and Information Technology
University of Queensland
Professor Zhiguo Yuan
Director, JC STEM Lab of Sustainable Urban Water Management
School of Energy and Environment
City University of Hong Kong
Previous Postdoc Research Fellow
Dr Zheng Zhang
Associate Professor
Harbin Institute of Technology (Shen Zhen)
Dr Jingjing Li
Feb 2019 - Apr 2019)
Professor
University of Electronic Science and Technology of China
Dr Weiqing Wang
Lecturer
Monash University
Dr Lei Zhu
Professor
Shandong Normal University, China
Dr Xiaoshuai Sun
Associate Professor
Xiamen University
Completed Research Students
Jiewei Cao
Senior Research Scientist
Bosch China
Chao Li
Buidu Research
Hongyun Cai
Senior Researcher at Tencent, China;
Advanced Digital Sciences Center
(ADSC), Singapore
Xuefei Li
Amazon Web Service, Canada
MPhil Eve Belisle
École Polytechnique de Montréal;
a Canadian curler and amateur
ornithologist from Montreal
Ziwei Wang
Postdoc
CSIRO
Luyao Liu
Yadan Luo
Postdoc
The University of Queensland
Jiasheng Duan
Ruihong Qiu
Postdoc
The University of Queensland
Zhi Chen
Postdoc
The University of Queensland
Yang Li
Research Fellow
CSIRO
Mr Pengfei Zhang