Conclusion

Although there have been many papers prototyping interesting applications of artificial intelligence in life science particularly biology and ecology, very few outline beginner-friendly approaches to classifying species. To address this issue, we have illustrated the tools, steps, and resources required to build a image classifier. Despite the existence and use of these tools, it is when citizens and scientists alike have access to readily available cost-effective and intuitive tools that domain ecologists can start utilizing the potential of these powerful algorithms to solve and discover otherwise challenging problems.  We hope this paper can help spur a new approaches to species classification. As a next step, we will use this methodology to build a more comprehensive sea star image classifier for the Western Coast of the United States. The classifier will be integrated into a phone application to allow citizen scientists to monitor both population counts and also sea star wasting disease.

Acknowledgments

This work was partially supported by the NSF traineeship: Quantitative & Evolutionary STEM Training (QUEST): An Integrative Training Program for Versatile STEM Professionals to Solve Environmental and Global Health Problems of NSF Award Number: 1735316 and Prof. Melissa Pespeni's lab start up grant at the University of Vermont. Dr. Easton White and  Prof. Nick Cheney provided input at the highest level. The workflow is adapted from the Fast.ai course run by Jeremy Howard and Rachel Thomas of the Data Science Institute of the University of San Fransisco.