San Francisco Estuary Institute: Analyzing Drone Data to Find Trash Fast

Each year, tons of trash sails down tributaries into the San Francisco Bay, polluting our environment, threatening wildlife, and contaminating our water. The San Francisco Estuary Institute has developed a new trash-detection method that uses a combination of drone imagery, captured from above public spaces around creeks and rivers, and a TensorFlow-based convolutional neural network (CNN), to expand the spatial range and number of times they can perform trash monitoring throughout the San Francisco Bay Area’s streams and rivers.
SFEI partnered with Kinetica and Oracle Cloud Infrastructure (OCI) to leverage GPU-based resources, including OCI's bare metal GPUs, to accelerate compute-heavy workloads on streaming and batch data, in search of greater performance. Kinetica, using OCI, can seamlessly move data back and forth from traditional big data architectures. The results were remarkable. A CNN training process that had taken more than ten days to run now takes minutes thanks to additional GPU horsepower and the Kinetica Active Analytics Platform. This allows the team to tune the model and perform comparison testing with faster iterations. You can watch a demo here • GPU-Accelerated Object...
We hope the project will deliver a trash detection model that can be applied broadly, throughout the Bay Area and beyond. Explore more examples of active analytics in action at www.kinetica.com

[@DrNadineGreinerPhD]

Tony Hale at SFEI does incredible work...bravo to those who partnered to put together this inspiring video. Thank you! I, for one, will be even more mindful.