Migration paths of the Carcharodon Carcharias (Great White Shark) are disproportionately altered by human-induced stressors on oceanic health, where ecosystem’s trophically cascade from rapidly changing environments. Current solutions to aid the erupting coastal food chain manually track white sharks (the primary indicator of ocean quality) with short term biomarkers. Unfortunately, modern solutions are expensive in mass, unreasonable in the long term, and harmful to white sharks. Thus, we present SharkHI: a deep learning system to detect, track, and predict shark health from video footage. Trained on the Ocean Predatory Lab (OPL) dataset, SharkHI provide a neural network model to predict white shark health conditions in a safe, inexpensive, and accurate manner, while proving a novel approach to calculate shark health more effectively. This system includes a convoluted neural network binary classifier to isolate viable shark image frames, and a pose detection network to track 5 key points for health condition calculation. SharkHI delivers a generalizable model proven to enable predictive conclusions on regional oceanic health conditions worldwide. Additionally, SharkHI explores an additional methodology using the center body of the white shark, without the need for the nose, pectoral/dorsal fins, or caudal keel in frame, vastly improving viable image frames for calculation. With SharkHI, an innocuous pipeline is prevalent to bridge foundational conclusions on necessary understandings of both shark and oceanic health globally.
Previous Article in event
Next Article in event
SharkHI: a Novel Pipeline for Health Index Calculation of Carcharodon Carcharias.
Published:
07 March 2025
by MDPI
in The 3rd International Electronic Conference on Animals
session Sustainable animal welfare, ethics and human-animal interactions
Abstract:
Keywords: Machine Learning, Great White Shark, Neural Networks, Carcharodon Carcharias
