Case Study: The Nature Conservancy improves fish detection and classification with Ciklum

Nature Conservancy - Customer Case Study

The Nature Conservancy, a US-based conservation charity, wanted a camera-based way to automatically detect and classify fish species caught by commercial fishing fleets. To address illegal, unreported, and unregulated fishing in the Western and Central Pacific, the organization launched a Fisheries Monitoring contest and worked with Ciklum’s computer vision and deep learning expertise.

Ciklum’s R&D team of three engineers built both a whole-image classification approach and a fish detection and classification algorithm using convolutional neural networks. The system achieved 95% localization precision and 80% recall for fish detection, 93% accuracy for fish species classification, and a log loss of 1.3 in the contest, close to the competition’s best score of 1.1.

The Nature Conservancy