Building on the strategic plans at UC San Diego and the Scripps Institution of Oceanography, the 100 Island Challenge takes an interdisciplinary approach to find innovative solutions to research questions and conservation issues. Working with partners in the computer science and engineering departments, we have developed a data collection and analytical pipeline that leverages advances in modern digital imaging and computing. The results of this partnership has been an unprecedented opportunity to collect and extract ecological information at previously unavailable rates. To accommodate these dramatic increases of data, we are also working with our partners to implement computer assisted classification approaches. Importantly, we are committed to fostering information exchange and data sharing.
IMAGING AND VISUALIZATION
In order to overcome the challenges of working in and collecting data from in subtidal environments, we have developed an imaging and visualization pipeline that fundamentally restructures the way that we can interact with coral reef environments. Using structure-from-motion software, we create highly detailed 3D models representing hundreds of square meters of coral reef habitats. When used with our custom visualization and analytical platform, these models become permanent detailed digital recreations – enabling the astute biologist, the curious onlooker or the concerned manager a limitless underwater experience. Ultimately, this imagery serves as a large-scale, photographic archive and rich data source for exploring coral reefs in critical and novel ways for decades to come.
Each 3D model contains an incredible amount of potential information, however current data extraction protocols require substantial human-driven data extraction efforts. In order to overcome this bottleneck we are working with our technical partners to implement machine learning and artificial intelligence to accelerate this workflow. While we believe that the expert biologist should never be removed from the equation, in order to scale our efforts we must augment the discerning human eye with the ever evolving capabilities of image recognition software.