CoralNet
A free, open platform that uses image recognition to help coral reef researchers annotate large volumes of underwater photos faster.
About
Researchers use CoralNet to label points on images of the seabed as, for example, hard coral, algae, or sand. Once a project has enough verified examples, a neural network can fully or partially pre-annotate new images. Recent evaluations show this can accelerate ecological monitoring, but performance varies by region, image quality, and species group. Experts therefore set thresholds and check uncertain labels.
AI factor
Manually labeling thousands of photo points consumes a lot of specialist time. The model takes over recognizable cases and routes doubtful ones to a researcher, keeping human attention focused on new or hard-to-distinguish organisms.
Sources
Inspiration sourced from AI For Good. Not an Arnaldus customer case.