←Back to World Data Ocean
•3 min read•from Frontiers in Marine Science | New and Recent Articles

RSF-YOLOv8: a re-parameterized multi-branch feature augmentation network for polarization-guided underwater object detection

Our take

Underwater object detection is critical for marine resource exploration, engineering inspection, and autonomous underwater vehicle (AUV) operations. Addressing challenges posed by turbid conditions and image degradation, researchers have developed RSF-YOLOv8, a novel model based on YOLOv8n that leverages polarization imaging. This innovation incorporates specialized modules for enhanced multi-scale feature extraction, improved target recognition, and precise localization, particularly for small objects.
RSF-YOLOv8: a re-parameterized multi-branch feature augmentation network for polarization-guided underwater object detection

The challenges of underwater object detection are increasingly vital as our reliance on autonomous underwater vehicles (AUVs) grows for tasks ranging from marine resource exploration to infrastructure inspection. Traditional optical imaging methods struggle in the complex underwater environment, plagued by scattering, attenuation, and distortion. This new research, detailing the RSF-YOLOv8 model, represents a significant step forward in addressing these limitations, leveraging polarization imaging to improve detection accuracy. It’s encouraging to see continued innovation in this area, particularly as we consider the broader implications for ocean monitoring and management, as highlighted in our recent piece First expedition Jaywun research vessel: assessment of microplastics from international waters, Spain to Abu Dhabi, U.A.E, where data collection is inherently reliant on robust detection capabilities. The need for precise and reliable data acquisition is also a recurring theme in discussions around the governance of marine protected areas, as explored in From spatial expansion to institutional coherence: governance pathways of marine protected areas in island blue economies, which emphasizes the importance of accurate environmental assessments for effective management.

The RSF-YOLOv8 model’s design is particularly noteworthy. The innovative modules—EfficientRep, EfficientSE, and FASFF4—demonstrate a thoughtful approach to optimizing performance for polarized images. The re-parameterized multi-branch convolution enhances feature extraction without increasing computational cost, crucial for deployment on resource-constrained AUVs. The focus on small target detection, a common pain point in underwater imaging, is also commendable. The reported improvements in mAP@[0.5:0.95] – a measure of strict localization accuracy – are particularly significant, indicating a tangible reduction in missed targets. These findings underscore the value of specialized architectures tailored to the unique characteristics of polarization data, moving beyond simply adapting existing models. The rigorous ablation studies and generalization experiments further strengthen the validity of the approach, showcasing its robustness across diverse underwater scenarios.

Beyond the technical details, the emphasis on real-time performance and embeddable deployment is a critical factor for practical application. The reported inference speed of 62.3 FPS on an RTX4060 GPU and 18.2 FPS on an Intel i9-13900HX CPU, coupled with the model’s relatively small size (3.1M parameters), positions it as a viable solution for AUVs. This aligns with the broader trend towards edge computing and on-board processing, reducing reliance on bandwidth-limited communication links and enabling more responsive autonomous operations. The validated, measurable improvements in precision and reduced false positives, even within a high-accuracy regime, demonstrate the potential for this technology to contribute meaningfully to improved data quality and operational efficiency in underwater environments. The work also adds to the growing body of research informing future MSc thesis ideas, as detailed in Looking for MSc Thesis Ideas in Hydrography, Geodesy & Geoinformatics, highlighting the ongoing need for skilled professionals in this field.

Looking ahead, a key question is how this technology can be integrated into broader ocean observation systems. Could RSF-YOLOv8, or similar polarization-guided detection models, form the basis for automated monitoring of critical habitats, detection of invasive species, or assessment of infrastructure integrity? The ability to accurately and efficiently identify and track objects in challenging underwater environments unlocks a wealth of possibilities for advancing our understanding of the ocean and managing its resources sustainably. Further research focusing on the model's performance in varying water conditions, alongside exploration of its applicability to different types of underwater targets, will be crucial in realizing its full potential.

Underwater object detection techniques for marine resource exploration, underwater engineering inspection, and AUV intelligent operations face three major demands: high precision, low latency, and embeddable deployment. In natural underwater environments, complex interferences such as water scattering, wavelength-selective attenuation, illumination distortion, and target occlusion commonly exist. As a result, traditional optical imaging methods often suffer from inherent drawbacks, including image blurring, low contrast, and edge degradation. Polarization imaging can separate scattered stray light and preserve the intrinsic polarization texture of targets. Therefore, it exhibits significant advantages in highly turbid and low-light underwater scenes. However, existing lightweight detection models are not structurally adapted to the unique features of polarization images, leading to weak multi-scale modeling capacity, suppression of weak target features, and high miss rates for small targets. To address these issues, a novel underwater object detection model termed RSF-YOLOv8 is proposed based on the YOLOv8n baseline. Three core innovative modules are designed to optimize the model: 1) An EfficientRep re-parameterized multi-branch convolution module is constructed. By combining multi-scale parallel convolutions with structural reparameterization, the multi-scale feature extraction capability is enhanced without increasing inference cost, thereby adapting to the large-scale span and irregular morphology of underwater targets. 2) An EfficientSE dual-pooling channel attention module is designed. Global average pooling and max pooling are fused to replace the traditional single-pooling SE structure so that target feature channels are adaptively enhanced while redundant background channels are suppressed, improving the recognition capability for low-contrast polarized targets. 3) A FASFF4 (Four-scale Adaptive Spatial Feature Fusion) module is established. Unlike the original PAN-FPN which only performs simple lateral connections across three scales(1/32, 1/16, 1/8) and lacks fine-grained 1/4 detail preservation, our FASFF4 introduces a dedicated 1/4-scale branch and a bidirectional cross-scale interaction pathway that explicitly transmits high-level semantics downward while compensating detailed edge cues upward through an additional bottom-up path. This four-scale bidirectional design alleviates feature dilution and localization offset for very small underwater target, which are inherently prone to vanishing in conventional three-scale fusion. Systematic experiments are conducted on a self-constructed polarized underwater organism dataset(PUOD) and four public/self-constructed generalization datasets(UDD, DUO, URPC2020, BMPD). The results show that, compared with the baseline YOLOv8n, the proposed RSF-YOLOv8 achieves a precision of 0.9394(from 0.9299), an mAP@0.5 of 0.9939(from 0.9839), and a mAP@[0.5:0.95] of 0.7125(from 0.6792) on the PUOD dataset. Although the 1.02% gains in precision and mAP@0.5 appear marginal in absolute terms, they represent a relative reduction of 13.4% and 6.2% in false positives and misclassifications, respectively, given that the baseline already operates in a high-accuracy regime(>0.98). More critically, the 4.90% improvement in mAP@[0.5:0.95] reflects a substantial enhancement in strict localization accuracy, which directly translates to fewer missed small targets in real AUV operations. Ablation studies validate the effectiveness and synergistic effects of each module, while generalization experiments demonstrate strong robustness across different underwater scenarios. This research provides a high-precision and real-time detection solution for AUVs equipped with polarization cameras, achieving an inference speed of 62.3FPS and 16.0ms single-image latency on RTX4060 GPU, as well as 18.2FPS on Intel i9-13900HX CPU after structural reparameterization. With only 3.1M parameters and 8.1 GFLOPs at inference, the model provides a computationally efficient architectural foundation that is well-suited for resource-constrained underwater platforms.

Read on the original site

Open the publisher's page for the full experience

View original article →

Tagged with

#research datasets