Monitoring Social Insect Activity with Minimal Human Supervision

Tarun Sharma, Julian M. Wagner, Sara Beery, William B. Dickson, Michael H. Dickinson, Joseph Parker; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2024, pp. 1244-1253

Tracking the behavior of animals and their group dynamics in nature offers a crucial look into the delicate ecological networks that compose wildlife diversity. The velvety tree ant (Liometopum occidentale) is an ecologically dominant ant species found in South Western North America; their extensive foraging activity shapes forest communities and their nests are a biodiversity hot-spot for a multitude of symbiotic invertebrates (myrmecophiles). Despite their vital role in the ecosystem their activity is largely unstudied. In this work we develop a multi-sensor camera trap named 'Ethocam' to capture ant behavioral patterns in the field and combine this technology with a computer vision approach to track colony activity in an undisturbed fashion. We demonstrate an accurate system for counting ants built with minimal human labeling. We show that L. occidentale activity drops rapidly through the morning and study the effect of environmental conditions on ant count. We also report the occurrences of the ants' interactions with other invertebrates in our camera trap data. Together these findings demonstrate the potential of our system to capture the behavior of Liometopum occidentale as well as its complex associations with various local species including symbionts potentially at landscape scale. Our study provides proof of concept for the promise of low-cost remote monitoring of social insect populations.