Understanding the Model for Image Classification (Beta)
How the model works
The MERMAID AI: Image Classification (Beta) model was developed in collaboration with CoralNet, using its pyspacer framework for training deep learning models for image classification. Details of the pyspacer framework can be found in the following publication:
Chen, Q., Beijbom, O., Chan, S., Bouwmeester, J., & Kriegman, D. (2021). A New Deep Learning Engine for CoralNet. In Proceedings of the International Conference on Computer Vision (ICCV) Workshops. |
The MERMAID AI model was trained using a diverse set of publicly available CoralNet images. By integrating the majority of public CoralNet training data into a single generalized model, the MERMAID AI model makes reasonable inferences for a diverse set of images.
You can review the performance results for the MERMAID AI model in this GitHub documentation.
MERMAID is committed to the continuous improvement and refinement of this model.
Citation Suggestion |
Benthic attribute labels used in the model
The image classification model is currently trained on both top-level benthic categories and coral genera with growth forms, listed below. MERMAID aims to continuously improve and expand the benthic attributes used in the training and model classifications. We will continue to update this list, as image classifications for more benthic attributes are added by MERMAID users.
Top-Level Benthic Categories:
Coral Genera and Growth Forms:
