Maseeh Summer Undergraduate Research Experience

Funding Source

Undergraduate Research and Mentoring Program (URMP)

Document Type

Poster

Publication Date

2025

Abstract

Knowing how many Prairie Falcon nestlings survive each day is an important part of monitoring their populations. Normally, this is done manually by looking at nest photos and tallying live chicks each day. For this project, we teamed up with the Quantitative Conservation Lab at Boise State University to see if we could automate that process. To tackle this problem, we fined tuned a vision language model, essentially chatGPT processing/outputting images. We tested 3 approaches and determined that Single-term prompts were the most reliable overall, with no false negatives and fewer false positives. The two strongest terms were “baby chick” and “nestling falcon.”

Persistent Identifier

https://archives.pdx.edu/ds/psu/45280

Download

Share

COinS