My research interests primarily focus on computational approaches to visual perception, including animal, human, and machine vision. Trained as a zoologist and experimental psychologist, I am particularly passionate about how visual scenes can be “understood” using computers and what comparisons can be drawn with biological visual systems. Understanding vision can help us to generate a positive impact on the world, for example, automatic disease detection systems to improve animal welfare, providing a better museum experience for visitors, or raising awareness of how the colours of animals work. 

As a University Enterprise Fellow, I work on translating research in automatic tracking of racehorses to a commercial product that will enable identifying problems automatically at the earliest stage possible.

Projects I work on:

Automatic disease detection and monitoring

I lead and co-lead several projects that are aimed to develop automated systems to detect and monitor disease in various species, including 

• Bovine respiratory disease complex in calves
• Behaviour classification in racehorses
• Endoscopic analysis of upper airway function in racehorses
• Heart rate estimation from video footage in dogs
• Quantifying mortality of hens across cage-free systems

These projects use artificial intelligence techniques, coupled with visible-range and thermal cameras, to identify pathologies or changes in behaviour through automatic tracking. I design and build sensory platforms, establish efficient data transfer protocols and implement machine learning techniques (e.g. deep neural networks) to analyse complex, multisensory data.

Camouflage

I have been involved with camouflage since my PhD and continue to run projects on animal colouration. For instance, why are tigers not green? Hiding in green vegetation in bright orange brown fur clearly does not sound like an optimal solution. Other animals, such as the primary prey of tigers, have different colour vision to us humans which makes these mighty felines look different to them. Our work supports the notion that when considering the optimal colours for a given environment and function (e.g. concealment or signalling), the visual system of observers has a significant effect. This research has led to numerous media collaborations and I was fortunate to work with Liz Bonnin, Steve Backshall and Sir David Attenborough.

As a member of the CamoLab, I am part of the team developing a toolkit to establish the best (or worst) camouflage for any object in any environment for any viewer using deep neural networks. A paper on establishing optimal colours and textures for concealing and visibility can be found here. This work has been further expanded to entirely synthetic (i.e. artificial) prey and predators: Our team has demonstrated how Generative Adversarial Networks, deep networks engaged in competition, can be utilised to simulate an evolutionary arms-race.

I also continue to engage with the research topic of my PhD: this work focused on how military camouflage uniform patterns evolved since the early 20th century. The project uses methods from computer vision to establish similarity metrics between patterns, and phylogenetics to model how patterns of allied and hostile countries have influenced each other’s designs.

Understanding how people engage with art and spaces

This project examines how we can predict preference to visual artworks by utilising eye movements of observers and paintings generated by deep neural networks. Gaining insight into how people perceive art could also lead to more inclusive environments and our team is using virtual reality and wearables recording physiological measurements to establish how participants engage with various spaces.