Multi-label classification for multi-temporal, multi-spatial coral reef condition monitoring using vision foundation model with adapter learning

2025-03-29

AI Summary

To address the low efficiency, high computational cost, and poor generalization of multi-label classification in multi-temporal, multi-spatial-scale coral reef health monitoring, this paper proposes a lightweight multi-label classification method leveraging the DINOv2 vision foundation model and Low-Rank Adaptation (LoRA) adapters. To our knowledge, this is the first application of LoRA to coral reef image analysis. Evaluated on standardized underwater images collected across 15 dive sites at Koh Tao, Thailand, the method achieves an accuracy of 64.77%, outperforming the best conventional model by 4.43 percentage points. Trainable parameters are reduced from 1.1 billion to 5.91 million—a 99.5% compression—enabling efficient, low-carbon deployment. Moreover, the approach demonstrates strong cross-seasonal and cross-site generalization, effectively balancing classification accuracy, inference efficiency, and sustainability requirements for large-scale reef monitoring.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Data Mining & Knowledge Management: Mining of Visual, Multimedia & Multimodal Data

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Coral reef ecosystems provide essential ecosystem services, but face significant threats from climate change and human activities. Although advances in deep learning have enabled automatic classification of coral reef conditions, conventional deep models struggle to achieve high performance when processing complex underwater ecological images. Vision foundation models, known for their high accuracy and cross-domain generalizability, offer promising solutions. However, fine-tuning these models requires substantial computational resources and results in high carbon emissions. To address these challenges, adapter learning methods such as Low-Rank Adaptation (LoRA) have emerged as a solution. This study introduces an approach integrating the DINOv2 vision foundation model with the LoRA fine-tuning method. The approach leverages multi-temporal field images collected through underwater surveys at 15 dive sites at Koh Tao, Thailand, with all images labeled according to universal standards used in citizen science-based conservation programs. The experimental results demonstrate that the DINOv2-LoRA model achieved superior accuracy, with a match ratio of 64.77%, compared to 60.34% achieved by the best conventional model. Furthermore, incorporating LoRA reduced the trainable parameters from 1,100M to 5.91M. Transfer learning experiments conducted under different temporal and spatial settings highlight the exceptional generalizability of DINOv2-LoRA across different seasons and sites. This study is the first to explore the efficient adaptation of foundation models for multi-label classification of coral reef conditions under multi-temporal and multi-spatial settings. The proposed method advances the classification of coral reef conditions and provides a tool for monitoring, conserving, and managing coral reef ecosystems.

Problem

• Improves coral reef condition classification using vision foundation models
• Reduces computational resources with adapter learning (LoRA)
• Enhances model generalizability across temporal and spatial settings

Innovation

• DINOv2 vision foundation model integration
• LoRA fine-tuning for efficiency
• Multi-temporal multi-spatial coral monitoring