A selection of some of the interesting projects that I’ve been involved in over the past couple of years.
You can find those with a webapp component on the web apps page.
15cBOOKTRADE Visualisation Suite
The 15cBOOKTRADE project looks at material evidence of early 15th century printed books (called incunabula) in order to find out more about their distribution and usage. The project comprises of many researchers and institutions around the world working to record and analyse this information.
We developed a visual analytical tool for the scholars called 15cV. The tool allows for complex queries to be constructed to the main evidence database and the spatiotemporal visualisation of the results, and was was developed using D3.js, React, and a Python-based backend utilising Tornado.
I have written a blog post detailing how the Google Earth components of the video were created.
Visual Multiplexing with Cardiovascular MRI Case Study
This work was developed in conjunction with cardiologist Dr. Cameron Holloway, who was at the time at the John Radcliffe Hospital, Oxford. We focused on the problem of providing more advanced visualisation tools for reviewing temporal CMR imagery, allowing the user to augment the imagery with a selection of glyph-based ‘probes’ that convey computed properties such as stress back to the user.
Our paper Visualizing Cardiovascular Magnetic Resonance (CMR) imagery: challenges and opportunities summarises our investigations into the role of visualisation in this area, and what opportunities exist for advancement.
The work eventually led to the development of a more general visualisation work, Visual Multiplexing, which addresses the problem of overlaying multiple pieces of information in the same screen areas at once and how we can reason about such occurrences.
VIBE: Escape Spreadsheet Hell
During my time as an RA at Swansea University, we run an incredibly interesting study into how SMEs throughout Wales were utilising data management and visualisation techniques. What we found was worse than we had feared: the majority of businesses seem to run almost entirely from Microsoft Excel spreadsheets. They are built upon over years by an employee who quickly becomes the ‘Excel guru’ and then inevitably leaves, and then left to grow more complex, with hack after hack introducing endless macro bugs. On top of this, they are emailed around to such a degree that nobody is sure who currently owns the more accurate version.
We looked at this as our chance to do some good. Our small team visited 20 during the course of the project. Each time, we would work with the company to look at their data problems, introduce them to fresh ideas on how to improve their data management practices to lead them in a more positive direction, and also introduce them to the power of basic data visualisation technologies for providing insight and saving time in data analysis.
LiveLayer: Real-time Projection of Traffic Video onto Geographical Maps
LiveLayer was developed as a prototype tool to answer the question “how feasible is it to show real traffic video feeds on something like Google Maps?”. In this prototype, we incorporated multiple pre-recorded traffic feeds, and developed a straightforward method for allowing the camera’s homography within the geographical map to be easily defined by simply placing coloured flags within the imagery. The homography transform was then applied by the GPU, along with a Gaussian Mixture Model for background subtraction to overlay the video onto the map.
Full details of the system can be obtained from our paper LiveLayer: Real-time Projection of Traffic Video onto Geographical Maps, or our more technical followup on the problem of correcting the ‘stretched’ appearance that vehicles can take on during the homography transform: Vehicle object retargeting from dynamic traffic videos for real-time visualisation.
Glyph-Based Video Visualisation for Semen Analysis
Our work with domain experts from reproductive sciences resulted in what is probably the most information-dense glyph ever devised. The general movement of sperm cells is something that has been studied for some time due it its importance in the determination of motility. In addition, the complex morphological properties of the flagella (the ‘tails’ of the cells) have recently been studied in order to provide richer data.
Our paper, Glyph-Based Video Visualization for Semen Analysis, was presented at VIS 2014 in Paris. Its contributions were twofold: (a) a computer vision technique for extracting the cells and their flagella from swimming sequences; and (b) a multidimensional glyph encoding some 20 dimensions of a cell’s properties into an image. Our glyph encoding effectively reduces the burden of watching animations and allows for samples to be compared.
I developed an accompanying web application, Sperm Zoo to allow for conference attendees to experiment with modifying the properties of a cell and viewing the resulting glyph representation.
V2: A Video Surveillance System
The visualisation team at Swansea, headed by Professor Min Chen, created the earliest video visualisation methods for summarising video sequences into a single image. The advantages of such a technique are enormous: the burden of watching long video sequences in a surveillance setting is minimised, with changes in the scene summarised using change detection filters in a visualisation comparable to direct volume rendering.