UKANFormer: Noise-Robust Semantic Segmentation for Coral Reef Mapping via a Kolmogorov-Arnold Network-Transformer Hybrid

2025-10-19

AI Summary

Existing global coral reef mapping products (e.g., Allen Coral Atlas) suffer from insufficient spatial accuracy and semantic consistency, hindering fine-grained boundary delineation and effective conservation. To address this, we propose GL-Trans—a novel semantic segmentation module designed for high-precision coral reef mapping under noisy label supervision, challenging the conventional paradigm that “annotation quality dictates model performance ceiling.” Built upon the UKAN architecture, GL-Trans integrates Kolmogorov–Arnold networks with Transformers and introduces a global–local attention mechanism within the decoder to jointly model long-range semantic context and fine-scale edge structures. Evaluated on real-world noisy-label data, our method achieves 67.00% IoU for coral classes and 83.98% pixel accuracy—surpassing both baseline models and the original noisy annotations in both quantitative metrics and visual boundary fidelity.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Coral reefs are vital yet fragile ecosystems that require accurate large-scale mapping for effective conservation. Although global products such as the Allen Coral Atlas provide unprecedented coverage of global coral reef distri-bution, their predictions are frequently limited in spatial precision and semantic consistency, especially in regions requiring fine-grained boundary delineation. To address these challenges, we propose UKANFormer, a novel se-mantic segmentation model designed to achieve high-precision mapping under noisy supervision derived from Allen Coral Atlas. Building upon the UKAN architecture, UKANFormer incorporates a Global-Local Transformer (GL-Trans) block in the decoder, enabling the extraction of both global semantic structures and local boundary details. In experiments, UKANFormer achieved a coral-class IoU of 67.00% and pixel accuracy of 83.98%, outperforming conventional baselines under the same noisy labels setting. Remarkably, the model produces predictions that are visually and structurally more accurate than the noisy labels used for training. These results challenge the notion that data quality directly limits model performance, showing that architectural design can mitigate label noise and sup-port scalable mapping under imperfect supervision. UKANFormer provides a foundation for ecological monitoring where reliable labels are scarce.

Problem

• Achieving high-precision coral reef mapping using noisy supervision data
• Improving spatial precision and semantic consistency in segmentation
• Mitigating label noise effects through hybrid network architecture design

Innovation

• Hybrid Kolmogorov-Arnold Network-Transformer architecture
• Global-Local Transformer block for semantic and boundary details
• Noise-robust design mitigates imperfect supervision limitations