Applying machine learning tools to determine the fine-scale distribution of copepods and their siphonophore predators in the Gulf of Maine
Abstract/Description/Artist Statement
Using a shadowgraph camera, we imaged copepods and one of their main predators, the siphonophore Nanomia cara (a gelatinous zooplankton ) in the Gulf of Maine. Siphonophores are very important for this ocean ecosystem because they help to control copepod populations and also impact the local fishery as they feed on fish larvae. Preliminary observations indicate a highly patchy environment, where siphonophores were concentrated in only a few stations. The copepods were more abundant and spread out more evenly compared to the siphonophores that were only seen in offshore stations at depths of 6m and below. A machine learning tool (Faster R-CNN model) was trained and used to detect and classify different marine particles captured by the camera. This model helped to identify the abundances of copepods and siphonophores in different areas reliably and much faster than a human.
Faculty Advisor/Mentor
Alexander Bochdansky
Faculty Advisor/Mentor Email
ABochdan@odu.edu
Faculty Advisor/Mentor Department
Ocean & Earth Sciences
College/School Affiliation
College of Sciences
Student Level Group
Undergraduate
Presentation Type
Poster
Applying machine learning tools to determine the fine-scale distribution of copepods and their siphonophore predators in the Gulf of Maine
Using a shadowgraph camera, we imaged copepods and one of their main predators, the siphonophore Nanomia cara (a gelatinous zooplankton ) in the Gulf of Maine. Siphonophores are very important for this ocean ecosystem because they help to control copepod populations and also impact the local fishery as they feed on fish larvae. Preliminary observations indicate a highly patchy environment, where siphonophores were concentrated in only a few stations. The copepods were more abundant and spread out more evenly compared to the siphonophores that were only seen in offshore stations at depths of 6m and below. A machine learning tool (Faster R-CNN model) was trained and used to detect and classify different marine particles captured by the camera. This model helped to identify the abundances of copepods and siphonophores in different areas reliably and much faster than a human.