Dell Technologies has announced a new deep learning technology model, launched in partnership with Australia-based conservation organisation Citizens of The Great Barrier Reef. The model will allow global citizen scientists to more quickly and accurately analyse reconnaissance images collected from the Great Barrier Reef during the next phase of the Great Reef Census (GRC).
The new Dell deep learning model will help better inform conservation efforts for the Great Barrier Reef, one of the world’s greatest natural wonders.
A previously implemented Dell edge solution deployed on watercraft automatically uploads data directly to the deep learning model via a mobile network for real-time image capture. This will enhance the capabilities of the GRC by speeding image analysis that previously solely relied on human volunteers – allowing citizen scientists to support prompt recovery efforts in areas that need it the most and during critical times of the year, such as the annual spawning season.
The deep learning analysis now takes less than one minute per photo, compared to seven or eight minutes in previous census phases. While it took 1,516 hours to review 13,000 images in the first GRC, the new model can analyse the same data set in less than 200 hours.
The initiative aligns to Dell Technologies’ environmental, social and governance (ESG) ambitions to advance sustainability, by creating technology that drives progress and working with customers, partners, suppliers and communities to enable climate action.
The GRC is a partnership across Asia Pacific and Japan (APJ) combining the expertise of the Citizens of the Great Barrier Reef team, Dell, researchers from University of Queensland (UQ) and James Cook University (JCU), Sahaj Software Solutions and citizen scientists.
Dell also worked with its data science team in Singapore to continually refine and carry out extensive community testing of the selected deep learning model to ensure that benchmarking standards were met.
Speaking exclusively to BW Businessworld, Dell Technologies said that its has developed the deep learning model for GRC over last five months by model by prioritising coral categories, understanding the dataset, cleaning the dataset, building hundreds of models with different model architectures and hyperparameters.
“This deep learning still relies on human input: someone on the reef takes a photo, and the deep learning model defines the coral outline and health status; then, someone else can check the outline and confirm the label before the data is sent to the University of Queensland,” explained Aruna Kolluru, Chief Technologist, Emerging Technology, Dell Technologies, Asia Pacific and Japan.
For this project, Dell Technologies worked with the University of Queensland to understand the 144 categories of reef organisms so that the deep learning could divide them into categories, starting at 13 categories and through an interactive learning process, into 11, and finally six to make it easier for humans to identify.
“Photos from the first and second phase of The Great Reef Census were used to train the model to segment the images for six different coral categories, along with the percentage of the cover for these categories,” Kolluru added.
BW Businessworld asked Dell Technologies team if the learnings from GRC would be being applied to the coral reef system in the Andaman and Nicobar Islands’ or in the vicinity of the Indian subcontinent. “Currently, no. However, everything that Citizens of the Great Barrier Reef is doing is designed to be scalable and sharable so it could be in future,” answered Kolluru.