PNNL scientists and engineers developed EyeSea, a software tool to automate underwater video footage. EyeSea uses a machine learning visual algorithm to flag when any aquatic species enters the camera frame, which is positioned to monitor a marine hydrokinetic (MHK) turbine.
EyeSea uses machine learning to automate the analysis of MHK turbines for scientists. EyeSea runs while a camera maintains view of an MHK, continuously flagging moments when marine animals appear in frame. This is done autonomously, so that scientists can review the flagged footage later and not sift through hours waiting for animals to appear. EyeSea’s initial test over two months required scientists to analyze 43 hours of footage and observe 20 fish interactions with no injuries. Scientists assessed that EyeSea has an accuracy of 85 percent, which they are improving through algorithm refinement.
EyeSea provides a way to accelerate the understanding the environmental of MHK turbines by minimizing the work scientists have to do. Scientists only have to analyze and observe the video segments flagged by EyeSea, ignoring the rest of the video which has no activity. MHK turbines have the capability to provide over 10% of the electricity demand of Pacific states’ but require a thorough environmental analysis prior deployment.