In the summer of 2023, Dr. Wolenski and the DeVision team stepped up to the task of aiding LSU's AGGRC lab in their challenge of counting frog eggs. They used machine learning to segment the eggs and count them. They created a Stardist-based model by annotating 180 images and predicted eggs with 99.3% accuracy!
Our team resumed this project with the objective of minimizing the data required to replicate the summer's results. We employed a bottom-up approach by examining egg ranges, categorizing them into groups like 10s, 20s, and so on, up to the 70s. The performance in the lower ranges (10s to 50s) wasn't as successful, but the 60s and 70s ranges yielded slightly better results than those from the summer. Impressively, our team managed to reduce the total number of images required to annotate from 180 to about 20.
Dr. Peter Wolenski wolenski@math.lsu.edu
Dr. Nadejda Drenska ndrenska@lsu.edu
Iswarya Sitiraju, Gowri Priya Sunkara
Ibrahem AlJabea, Christian Ennis, Monika Pandey
Finn Phayer, Skylar Wilson
Director Terrance Tiersch
Assistant Director Prof. Yue Liu
Post Doc Dr. Jack Koch
Research Assistant: Carl Anderson
https://www.math.lsu.edu/courses/capstone_course
