Using computer vision methods to quantify pig Defence Cascade responses

Whether animals are in a positive or negative affective (emotional) state is key to their welfare. Assessing such states is therefore an important goal in animal welfare science. When presented with unexpected stimuli, the Defence Cascade (DC) response of a startle and freeze, is a potential indicator of affective state. Both humans and rodents in a negative affective state show increased startle magnitude and freeze duration. Pigs are known to show a clear DC response, but it would not be possible to take measurements on farms using the main approaches applied to humans (sensors recording eye blinks) or rodents (force plates). This paper aimed to evaluate whether DC responses in pigs could instead be accurately measured using computer vision. The experiment involved recording 280 induced DC responses in pigs who were standing on a force plate, enabling us to compare the sparse feature tracking computer vision image analysis with a trained observer analysing the videos by eye, measures from the load platform, a Kinect depth camera and for a subset also Kinematic data. Image analysis data were strongly positively correlated with the analysis by the trained observer and all other measures of the DC responses. Computer vision image analysis is thus a practical approach to measuring pig DC responses, which with further work could be applied under field conditions as a potential measure of affective state and thus welfare.

Statham, P., Hannuna, S., Jones, S. et al. Quantifying defence cascade responses as indicators of pig affect and welfare using computer vision methods. Sci Rep 10, 8933 (2020).

Full article / DOI can be found here.