Phase 2 Response for Integrated Time Series

Feedback: "It is our understanding that Edgington/Cline et al. have made good progress in using machine learning to identify a number of animals in video feeds that were highlighted as key species in earlier versions of this proposal. What has become of that advancement?"

The machine learning work targeting the ITS key species did indeed show very promising results. Duane and Danelle have outlined the detailed results of the ITS key species exploration work in the 2018 and 2019 proposals showing the results of their work and collaborations.  This is an excerpt from that text:

"Classification accuracy was between 62 and 100 percent for key ITS benthic and midwater using training images from the Deep-Sea-Guide and Dallas Hollis 2016 summer intern data"

With this success there are plans to examine the possible integration of these automated methods with the work being done in the Smith Lab. The new Smith Lab postdoc, Jen Durden, is interested in machine learning and will likely be working with Duane and Danelle towards partially automating the annotation of benthic transects.  In addition, the models and lessons learned from the ITS key species exploration will certainly be utilized in the workplan being articulated in the Video TAG proposal.