What question did this study set out to answer?

To evaluate the transferability of AI models trained for detecting cold-water coral morphology across different underwater imaging platforms and locations.

April 1, 2026Ecological InformaticsOpen Access

Testing the transferability of AI models for cold-water coral detection and classification

View Full Paper

Ask AI

Bookmark

To evaluate the transferability of AI models trained for detecting cold-water coral morphology across different underwater imaging platforms and locations.
Trained YOLOv5 models on coral imagery collected by ROV ISIS in the UK.
Tested model performance using independent datasets from Tropic Seamount and ROV Holland I.
Evaluated model performance based on recall of human annotations across different morphological groups.
Model performance recalled only 23% to 34% of human annotations when evaluated with independent datasets.
Moderate recall of 60% was noted against the validation dataset with variability across groups.
Poor transferability attributed to high shape variability and taxonomic mismatches between datasets.

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Tests AI models for coral detection and classification in different underwater environments, suggesting improvements are needed.

Cite This Study

Graves et al. (2026) studied this question.

synapsesocial.com/papers/69cd7a915652765b073a7c61 https://doi.org/10.1016/j.ecoinf.2026.103749

View Full Paper

Ask AI

Bookmark

Testing the transferability of AI models for cold-water coral detection and classification

Key Points

Discussion

Implication

Cite This Study

Testing the transferability of AI models for cold-water coral detection and classification

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider