September 2024

Deep learning for multilabel classification of coral reef conditions in the Indo-Pacific using underwater photo transect method

Research Overview Because coral reef ecosystems face human activities and climate change threats, coral reef conservation programs are implemented worldwide. Monitoring coral health provides references for guiding conservation activities. However, current labor-intensive methods result in a backlog of unsorted images, highlighting the need for automated classification. Few studies have simultaneously utilized accurate labels along with updated algorithms and datasets. This study aimed to create a dataset representing common coral reef conditions and associated stressors in the Indo-Pacific. Concurrently, it assessed existing classification algorithms and proposed a new multilabel method for automatically detecting coral reef conditions and extracting ecological information. A dataset containing over 20,000 high-resolution coral images of different health […]