Deep learning for multi-label classification of coral conditions in the Indo-Pacific via underwater photogrammetry

Since coral reef ecosystems face threats from human activities and climate change, coral 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 annotations along with updated algorithms and datasets. This study aimed to create a dataset representing common coral conditions and associated stressors in the Indo-Pacific. Concurrently, it assessed existing classification algorithms and proposed a new multi-label method for automatically detecting coral conditions and extracting ecological information. A dataset containing over 20,000 high-resolution coral images of different health conditions and stressors was constructed based on the field survey. Seven representative deep learning architectures were tested on this dataset, and their performance was quantitatively evaluated using the F1 metric and the match ratio. Based on this evaluation, a new method utilizing the ensemble learning approach was proposed. The proposed method accurately classified coral conditions as healthy, compromised, dead, and rubble; it also identified corresponding stressors, including competition, disease, predation, and physical issues. This method can help develop the coral image archive, guide conservation activities, and provide references for decision-making for reef managers and conservationists. The proposed ensemble learning approach outperforms others on the dataset, showing State-Of-The-Art (SOTA) performance. Future research should improve its generalizability and accuracy to support global coral conservation efforts.

💡 Research Summary

Coral reef ecosystems are currently facing unprecedented threats from anthropogenic activities and global climate change, making effective monitoring a cornerstone of conservation efforts. However, traditional monitoring methods rely heavily on manual inspection of underwater imagery, a labor-intensive process that leads to significant backlogs and delays in ecological assessment. To address this bottleneck, this study presents an automated approach using deep learning for the multi-label classification of coral conditions in the Indo-Pacific region.

The researchers first addressed the critical issue of data scarcity and outdated datasets by constructing a massive, high-resolution dataset comprising over 20,000 images. This dataset is uniquely comprehensive, capturing not only the primary health states of corals—categorized as healthy, compromised, dead, and rubble—but also the specific environmental stressors such as competition, disease, predation, and physical damage. This multi-label approach allows for a much more nuanced understanding of reef health than traditional single-label classification.

In terms of methodology, the study rigorously evaluated seven representative deep learning architectures to identify the most effective framework for this complex task. The core innovation lies in the proposal of a new ensemble learning method. By integrating the predictions of multiple architectures, the ensemble approach mitigates the individual weaknesses of single models, leading to superior performance in identifying overlapping ecological features. The performance of the proposed model was quantitatively validated using the F1-score and the match ratio, demonstrating that the ensemble method achieves State-Of-The-Art (SOTA) performance.

The implications of this research are profound for marine biology and environmental management. The ability to automatically extract detailed ecological information from large-scale underwater photogrammetry enables the rapid development of coral image archives and provides reef managers with actionable, real-time data. Such automated insights are crucial for guiding conservation interventions and making informed decisions regarding reef protection. While the current model shows exceptional results in the Indo-Pacific, the study concludes that future research should focus on enhancing the generalizability and accuracy of the model to support global-scale coral conservation efforts across diverse marine environments.