By Autumn Lee, Mount Holyoke College rising senior, GEMM Lab REU Intern 2023
Hello! My name is Autumn Lee, and I am a GEMM lab REU student this summer being mentored by Allison Dawn and Dr. Leigh Torres! I am a rising senior at Mount Holyoke College studying Neuroscience and Behavior, focusing on coastal and marine science. It has been a pleasure working with the GEMM Lab this summer, and I have enjoyed learning more about the field of research before I graduate.
As part of the research experience for undergraduates (REU) program, I am doing an independent project this summer in addition to our intense fieldwork for the TOPAZ project. I am working with the CamDo underwater video data that the GEMM Lab has collected since 2020. You can read Allison’s recent blog post to learn more about our CamDo underwater housings. Over the previous seasons, scuba divers have deployed our CamDo’s in our two study sites near Port Orford Titchener Cove and Mill Rocks on a weekly schedule of collection and redeployment. My project focuses on developing a methodology for examining the interactions between zooplankton prey and marine predators, and to quantify zooplankton density from the swarms seen on camera. Even though I hope my project’s success will contribute to the field, embarking on new method protocols always carries a risk of failure. Science tends to focus on successes; only in the footnotes do we hear about failures, wrong turns, and forgotten ideas. However, failure is how research advances; and with scientists who are brave enough to take that first step and humble enough to accept and reflect on failure.
Figure 1: Team prepping CamDo setup for deployment
In the past, I have learned to troubleshoot computer software and lab equipment. However, there were already protocols in place, and my research contributions were part of another student’s pre-defined project. Unlike my previous research experience, for my REU project, I had to learn how to use unfamiliar software, set achievable goals, overcome obstacles, and devise a plan to accomplish them without relying on a team of peers. This is a project Allison and I have been working on together outside of field work, but we have not been without support. Both Victoria Hermanson, a Biological Science Aid with the Antarctic Ecosystem Research Division, and Suzie Winquist, a graduate student at the Marine Mammal Institute, have inspired and guided us through using VIAME for our research questions.
Taking that leap into uncharted waters, we chose to work with two software programs that were new to me called VIAME (Video and Image Analytics for the Marine Environment) and ImageJ. Our goal was to utilize VIAME so that it could distinguish between zooplankton or predators in our CamDo videos (from the hundreds of unannotated frames) and then use ImageJ to quantify the density of zooplankton in those identified frames. Although it has been exciting to use this software that uses Artificial Intelligence (AI) to track and detect prey and predator interactions in video footage, we have encountered many challenges along the way. Within 10 weeks, we had to learn this new software, train it to identify zooplankton and predators, and calculate density using classified frames that we would train. When tackling such an ambitious project in a limited time frame, we expected some setbacks, and through the advice of experienced professionals and the support of Allison (as well as a healthy dose of self-determination), we were able to gain success by breaking down the project into smaller tasks and using trial and error to fix any issues that arose.
Figure 2: Photo of Allison and myself working together to problem solve a VIAME error
Although we have had some failures along the way, we have accomplished a lot, and I am eager to share some results with you. First, we developed and fine-tuned a workflow in VIAME to use AI to identify zooplankton prey and predators in our CamDo videos.
Figure 3: Screenshot of VIAME program that illustrates how we trained a model to identify zooplankton prey (yellow boxes) and fish predators (blue box) in the CamDo videos.
In addition, we implemented a workflow in ImageJ (another software program designed to process and analyze scientific images) to quantify zooplankton density from frames identified by VIAME with zooplankton. Even though it took a lot of trial and error, our primary objectives were met, and we learned a great deal for future GEMM projects.
Figure 4: An example processed output image depicting how ImageJ recognized bodies of zooplankton (black outlines) and counted individual zooplankton ( red dots).
While working on my independent project, I learned that an ability to troubleshoot software and data processing can apply to tricky field work situations as well. For instance, when we lost a weighted cage attachment that protects our RBR concerto sensor, we needed a temporary solution until the divers recovered our lost gear. So our team discussed a few different DIY options. After a frantic afternoon of trial and error, we ultimately decided on using a milk jug as a temporary cage. While it wasn’t the most glamorous solution, the GEMM lab is known to think outside the box as a fundamental part of both the fieldwork and research process.
Figure 5: Photo of Allison testing out our RBR milk jug temporary setup
I have found through this experience that sometimes it is more valuable to struggle and learn skills than to immediately succeed. I am hopeful that this lesson has prepared me for my future, and I couldn’t be more grateful. It has been an interesting summer for me as far as adapting to failures and embracing them. It was a difficult transition leaving my new friends at Hatfield in Newport where I spent my first 4 weeks and embracing an entirely different living dynamic here in Port Orford. With the field season and my research approaching its end, I realize how much I appreciate all the new people I have met here. Before this summer, I had not had many opportunities to interact with similar and enthusiastic marine scientists. Now I live and work with marine science mentors and peers in the field every day, which has been an invaluable experience, and I am grateful for the opportunity to learn from and interact with these inspiring people. It has been a meaningful summer, and I look forward to continuing to build relationships and learn from my failures during this next phase of my life.
Figure 6: Photo of Zoop Troop, from left to right Natalee, Autumn, Allison, Jonah, Aly
Did you enjoy this blog? Want to learn more about marine life, research, and
conservation? Subscribe to our blog and get a weekly message when we post a new
blog. Just add your name and email into the subscribe box below.