Automated Approaches to Bowhead Whale Identification

Author

Kacey La, Ursinus CollegeFollow
Sam Gregory, Ursinus CollegeFollow

Submission Date

7-29-2022

Document Type

Paper

Department

Computer Science

Faculty Mentor

Christopher Tralie

Second Faculty Mentor

Leslie New

Comments

Presented during the 24th Annual Summer Fellows Symposium, July 22, 2022 at Ursinus College.

Included as a supplemental file is a PowerPoint poster.

Project Description

This project aims to automate the identification of bowhead whales using convolutional neural networks. The initial neural network identifies key points to outline each whale and uses these points to divide each whale into three sub-sections: the fluke, the back, and the head. Upon segmenting the whale, each sub-section was used to identify individual bowhead whales through the white patterns and scarring on their backs. The results from each segment were then combined into a final classifier to identify bowhead whales.

Recommended Citation

La, Kacey and Gregory, Sam, "Automated Approaches to Bowhead Whale Identification" (2022). Computer Science Summer Fellows. 5.
https://digitalcommons.ursinus.edu/comp_sum/5