Visual modelling
I have always been fascinated by vision as a sense; it lets us inspect tiny things up close, or vast galaxies far away. Our subjective experience as humans is incredibly visual and vivid, giving us the false belief that our eyes are a faithful window to the world outside our bodies. But – spoiler – they are not! Our eyes are severely limited, only capturing a tiny proportion of the information available, and then processing it in lossy, limited ways. Other animals have vastly different eyes and brains, so their experiences will be wildly different to our own. Being able to see the world through other animal’s visual systems is therefore not only deeply thought provoking, it’s also vital for understanding many aspects of animals’ lives. So visual modelling is a major aspect of my work.
Colour vision
Attempting to objectively infer another animal’s subjective visual experience is a difficult task, but we can quite reliably model how the receptors in an animal’s eye respond to light. I have developed a range of tools for doing this with digital cameras (micaToolbox), and have created open-source spectroradiometers and imagers (hardware & software). This allows us to quantify the colours visible to another animal, including extending spectral sensitivity in to the ultraviolet range (a portion of the spectrum that we can’t see, but that many other animals can).
The next steps are much more messy. We need to convert the colours seen by different eyes into relevant colour spaces, and integrate colour with pattern (spatial vision) and/or time (motion vision).
Spatial vision
The spatial arrangement of colour patterns give rise to the most wonderful displays in nature, from vivid sexual signals and warning colours, to camouflage that matches the exact patterns and colours of an animal’s surrounds. Many displays will even have dual-functions, acting as signals up close, but blending in with the environment perfectly from a distance (e.g. the black-and-yellow warning stipes on a caterpillar will blend together and make a background-matching brown from a distance). I’ve developed a wide range of tools that can objectively measure colours and their spatial arrangements simultaneously (spatio-chromatic modelling). This includes the QCPA framework with a whole host of tools, GabRat for quantifying edge disruption, distance-dependent modelling, and various pattern analysis tools.
We like to imagine colours can be expressed in ‘colour spaces’, but colour appearance actually depends on spatial (and temporal) information too. For example, place a grey disk on a red-ish background and it will look green-shifted, now move the same grey disk to a green-ish background and it will look red-shifted. This effect is called simultaneous contrast, and we’ve been aware of it for about 1,000 years. But now if you step back or use smaller disks, and make the backgrounds more intense red and green, the disk’s perceived colours will flip the other way; the disk on the green background looking more green and vice-versa, called assimilation (or White’s illusion). Existing visual modelling likes to ignore these dramatic effects, however I have developed a comparatively low-level visual model (the spatio-chromatic bandwidth limited model) that can predict and quantify them.
Motion vision
Motion is very ‘easy’ for visual systems to detect; think of a camouflaged animal moving and suddenly being easy to spot. However, it’s much more difficult for visual system to convert this detected change in brightness into estimations of the speed and direction of a moving object. High contrast spots and stripes can interfere with the way the basic mechanisms of motion detection work. I have developed methods that simulate how ‘elementary motion detectors’ work, and have used these to show how butterfly wing markings blend with wing dynamics to confuse birds.
