Machine vision for assessment of enrichment items in Lesser Egyptian Jerboa (Jaculus jaculus)

Jerboas (family Dipodidae) present a compelling model to study biomechanics, as they are bipedal hopping rodents whose ricochetal escape responses involve unpredictable three-dimensional trajectories. However, these rapid maneuvers may make it difficult for observers to score behavior, as small lapses in attention may miss multiple behaviors. We hypothesize that machine vision will reach human-grade observations, with resistance to observer fatigue and drift. After a standardized orientation, we asked 9 observers to score videos of captive jerboa- achieving a mean accuracy of 71% and precision of 90% (12 minutes, 9 behaviors). We then trained a machine vision classifier with approximately 8 hours of annotated data to achieve an overall accuracy of 78% and precision of 92%, meeting our benchmarks and allowing for assessment of a larger dataset. This included a cross-over study wherein jerboa (n=8) were offered enrichment items for 2 weeks at a time. With these ethograms, multiple factors were examined- and fur thinning was noted to correlate with the behavior of grooming. Furthermore, it then appeared that socially based enrichment items decreased the incidence of behavioral grooming, suggesting the potential predisposition of grooming for maladaptive behaviorism in captive jerboa. By enhancing the speed and accuracy of ethogram-based observations, this machine vision classifier has the potential to enhance our ability to study behavior, optimize husbandry, and establish standard animal care protocols for novel model systems.

@inproceedings{boulanger2025machine,
  title={Machine vision for assessment of enrichment items in Lesser Egyptian Jerboa (Jaculus jaculus)},
  author={Boulanger, Matthew and Miyamae, Juri and Hish, Gerry and Moore, Talia},
  booktitle={INTEGRATIVE AND COMPARATIVE BIOLOGY},
  volume={65},
  pages={S61--S61},
  year={2025},
  organization={OXFORD UNIV PRESS INC JOURNALS DEPT, 2001 EVANS RD, CARY, NC 27513 USA}
}