O'Keeffe, J. | Search bioRxiv

bioRxiv2021

A Computational Model of Stereoscopic Prey Capture inPraying Mantises

We present a simple model which can account for the stereoscopic sensitivity of praying mantis predatory strikes. The model consists of a single "disparity sensor": a binocular neuron sensitive to stereoscopic disparity and thus to distance from the animal. The model is based closely on the known behavioural and neurophysiological properties of mantis stereopsis. The monocular inputs to the neuron reflect temporal change and are insensitive to contrast sign, making the sensor insensitive to interocular correlation. The monocular receptive fields have a excitatory centre and inhibitory surround, making them tuned to size. The disparity sensor combines inputs from the two eyes linearly, applies a threshold and then an exponent output nonlinearity. The activity of the sensor represents the model mantiss instantaneous probability of striking. We integrate this over the stimulus duration to obtain the expected number of strikes in response to moving targets with different stereoscopic distance, size and vertical disparity. We optimised the parameters of the model so as to bring its predictions into agreement with our empirical data on mean strike rate as a function of stimulus size and distance. The model proves capable of reproducing the relatively broad tuning to size and narrow tuning to stereoscopic distance seen in mantis striking behaviour. The model also displays realistic responses to vertical disparity. Most surprisingly, although the model has only a single centre-surround receptive field in each eye, it displays qualitatively the same interaction between size and distance as we observed in real mantids: the preferred size increases as prey distance increases beyond the preferred distance. We show that this occurs because of a stereoscopic "false match" between the leading edge of the stimulus in one eye and its trailing edge in the other; further work will be required to find whether such false matches occur in real mantises. This is the first image-computable model of insect stereopsis, and reproduces key features of both neurophysiology and striking behaviour.

neuroscience↗

bioRxiv2021

Face dissimilarity judgements are predicted by representational distance in deep neural networks and principal-component face space

Human vision is attuned to the subtle differences between individual faces. Yet we lack a quantitative way of predicting how similar two face images look, or whether they appear to show the same person. Principal-components-based 3D morphable models are widely used to generate stimuli in face perception research. These models capture the distribution of real human faces in terms of dimensions of physical shape and texture. How well does a "face space" defined to model the distribution of faces as an isotropic Gaussian explain human face perception? We designed a behavioural task to collect dissimilarity and same/different identity judgements for 232 pairs of realistic faces. The stimuli densely sampled geometric relationships in a face space derived from principal components of 3D shape and texture (Basel Face Model, BFM). We then compared a wide range of models in their ability to predict the data, including the BFM from which faces were generated, a 2D morphable model derived from face photographs, and image-computable models of visual perception. Euclidean distance in the BFM explained both similarity and identity judgements surprisingly well. In a comparison against 14 alternative models, we found that BFM distance was competitive with representational distances in state-of-the-art image-computable deep neural networks (DNNs), including a novel DNN trained on BFM identities. Models describing the distribution of facial features across individuals are not only useful tools for stimulus generation. They also capture important information about how faces are perceived, suggesting that human face representations are tuned to the statistical distribution of faces.

neuroscience↗