Comparing Web Diversity in the Spider Uloborus Diversus

Maya Filipovitz

Mentor: Dr. Gordus

Department of Biology at Johns Hopkins University

How behavior is generated, specific to individuals, is not well understood. Various methods used, in hopes of measuring similarities and differences between individuals, generally involve qualitative analysis, which is prone to human bias. In the Gordus Lab, spiders are used as model organisms to study behavior. Specifically, the lab examines various methods, with an emphasis on quantitative analysis, to determine how to best measure and observe spiders’ web making process, which can tell us a lot about behavior. In this study, we relied on video recordings of spider web making in real time to observe both the spider’s legs (50 Hz) and web (10 Hz) clearly visible. Using convolutional neural networks (CNNs), the legs were automatically tracked, while the final web structure was annotated by hand. With these annotations, the web geometry was analyzed using quantified metrics, such as line lengths and angles, which will allow for future comparisons to be made between webs and building individuality.