AI accuracy across marine-analysis tasks
Reported AI performance in marine monitoring: accuracy exceeds 90 percent across multiple tasks, with some applications reaching 98–99 percent.
99%
Whale species classification Fish species identification
98%
94%
Plankton classifiers
90%
Camera-trap image labelling
Turning data volume into insight The volume of marine data now being generated makes manual analysis increasingly impractical. AI is helping close that gap. Across global studies, machine-learning models have demonstrated strong performance in tasks such as species identification and behavioural analysis. For example, deep-learning tools have classified whale species from imagery with 98 percent accuracy and estimated body length within 5 percent of manual measurements.⁶ Fish species have been identified in video datasets with more than 94 percent accuracy,⁷ and plankton classifiers routinely exceed 90 percent
accuracy.⁸ In large camera-trap studies, AI has automatically labelled more than 99 percent of images, reducing manual labelling by more than 17,000 hours.⁹ Acoustic analysis is improving as well. NOAA researchers found that blending machine learning with synthetic data increased detection precision from 86 percent to 90 percent and recall from 88 percent to 93 percent. 10 But high performance does not always require deep learning. For well-defined animal calls such as pygmy blue whale signals, conventional template-matching techniques remain highly effective. When deep learning is used, adaptation to local conditions is essential. Transfer learning allows pre-trained models
Nexus Magazine | GHD | 9
Made with FlippingBook - professional solution for displaying marketing and sales documents online