Skip to main content
Entomology 2024 Main banner
Bayer Crop Science
Icon Legend
 
Oral Presentation
Oral Presentation
Poster Display
Poster Display
Student Competition
Student Competition
Ticketed Event
Ticketed Event
Back

Program Symposium

Publications

Artificial Intelligence (AI) in Entomology: An Aid to Publishing, Research, and Teaching

AI and modern CV in the insect laboratory: Discovering important behaviors in ants and deep faking wasps

Wednesday, November 13, 2024
11:25 AM – 11:45 AM MST
Location: Phoenix Convention Center, 121 C, PCC
Oral Presentation
Applications of AI in insect behavioral research have largely been improvements of classic methodologies. For example, AI has made tools for automatically tracking the position of insects far more efficient, but legacy approaches continue to dominate the processing of the resulting tracking data. Furthermore, the need for tracking seems inconsistent with capabilities of AI to reason over raw images. Additionally, the disruptive power of generative AI has not reached insect behavioral research. Here, we review two examples of novel methods in insect social behavior possible due to sophisticated AI and computer vision (CV) techniques. First, we present a deep neural network model that learns to identify colony status from short segments of raw video data and then visualizes what behavioral features it uses to make predictions. As a proof of concept, we applied this network to 20-day recordings of colonies of Harpegnathos saltator ants during reproductive hierarchy re-formation. Without tracking or human annotation, the trained model highlighted behaviors known to be important in the process and thus has the potential of finding novel individual-level behaviors that warrant more study. Next, we present early work in using motion-transfer models from computer vision to transfer the face of one paper wasp onto the video-recorded behaviors of another, a "deep fake" approach that can be used in comparative studies of individual recognition in Polistes hierarchy formation. These two approaches showcase how AI and CV can open up novel methodological approaches and potentially new research directions in insect behavior.