Madelyn Krischer1, Gabriel Mopps1, Jack Prior2, Dr. Patrick Schwing1
1- Eckerd College St. Petersburg, FL
2- National Oceanic and Atmospheric Administration Affiliate
This neural network is a pilot study focused on using a pipeline of deep learning programs in an effort to characterize and identify species of foraminifera. Foraminifera are single celled organisms useful for determining environmental health and measuring ecological and climatic change. The current methodology using foraminifera is labor-intensive, requiring not only specialized taxonomists, but also substantial amounts of time and financial commitments for researchers and organizations to utilize the data. The use of a neural network trained in specific hard-shelled taxa can be used to increase the throughput of samples or entirely automate the process if the system becomes completely trained. The neural network has been designed using the framework and interface of VIAME, an open-source program specializing in video and image analytics in marine environments originally designed to analyze underwater video and imagery for fisheries stock assessments. Future expectations for the program include the automation of planktic flux and the Ammonia/Elphidium (A/E) index calculations, which is an important biological indicator for hypoxia. These tools will be applied to assist in ongoing harmful algal bloom monitoring using both foraminifera. This program’s completion will increase the capability of any research groups in the Gulf of Mexico or Atlantic basin region using foraminifera, allowing for unparalleled automation and throughput with applications ranging from paleoclimate and geohazard work to pollution research.
For more information email: mekrischer@eckerd.edu