AquaFusionNet: A Deep Feature Fusion and Attention-Based Approach for Accurate Underwater Defect Detection

Abstract Underwater infrastructures such as pipelines, ship hulls, and offshore platforms are critical to marine operations but are highly vulnerable to biofouling, structural corrosion, and vegetation overgrowth, leading to increased maintenance costs and environmental hazards. However, visual inspection of these structures remains challenging due to low visibility, uneven lighting, and complex textured surfaces that limit the effectiveness of both traditional and purely deep learning-based approaches. In this work, we introduce AquaFusionNet, a hybrid defect classification framework that seamlessly integrates embeddings from four state-of-the-art pre-trained backbones (EfficientNet-B0, ResNet-50, SENet-50, and Vision Transformer) with complementary traditional descriptors including colour histograms, histogram of oriented gradients (HOG), local binary patterns (LBP), edge density, and gradient statistics via a trainable attention module. This attention mechanism dynamically weights each feature channel, allowing the model to emphasise the most informative cues while preserving fine-scale details under variable turbidity and illumination. We evaluate AquaFusionNet on a curated dataset of 2,228 underwater images spanning three defect categories, with 445 images held out for testing. Our model achieves 98.43% accuracy, 98.18% precision, 97.79% recall, a 96.06% intersection-over-union, and a 97.97% F $$_1$$ -score, outperforming eleven strong baselines, including ResNet-152 and EfficientNet-B0, by a substantial margin. These results demonstrate AquaFusionNet’s robustness and generalisability, paving the way for real-time, automated underwater inspection systems that can significantly enhance operational safety and reduce maintenance costs across marine industries. Graphic Abstract This graphical abstract provides a clear overview of the AquaFusionNet research, illustrating a streamlined workflow moving from data acquisition through final results and conclusions. The dataset comprises 2,228 underwater images categorized into biofouling, holes, and vegetation, sourced from the ICIP MVI-2024 challenge and split into training, validation, and testing subsets with a ratio of 60:20:20. The analysis phase emphasizes preprocessing steps combined with traditional descriptors, including color histograms, texture patterns, and gradient-based features, complementing the extracted deep features. The AquaFusionNet model uniquely integrates embeddings from four advanced, lightweight backbones, EfficientNet-B0, ResNet-50, SE-ResNet-50, and Vision Transformer (ViT-B/16) with traditional features, using a dynamic trainable attention mechanism. This attention module adaptively emphasizes the most informative features, which are then classified via a two-layer multilayer perceptron (MLP). Evaluation of the model reveals exceptional performance metrics, achieving an accuracy of 98.43%, an F1-score of 97.97%, and an intersection-over-union (IoU) of 96.06%, surpassing eleven state-of-the-art models on a test set of 445 images. The final conclusion highlights AquaFusionNet as an efficient, deployment-ready framework capable of robustly detecting underwater defects, significantly enhancing the safety and reducing the costs associated with marine inspections.