Abstract:Underwater object detection has been shown to exhibit significant potential for exploring underwater environments. However, underwater datasets often suffer from degeneration due to uneven underwater light distribution, complex underwater environment, and crowded underwater dynamic background. Thus, object detection performance would be degraded accordingly. In this paper, a large kernel convolutional object detection network based on self-attention and long-range relationship capture is proposed. Firstly, a hybrid dilated large kernel attention mechanism is proposed, which adopts the idea of hybrid dilated convolution and combines the advantages of large kernel attention mechanism and self-attention. This attention mechanism can avoid self-attention defects while achieving self-attention adaptiveness and long-range relevance. Secondly, a feature enhancement block called residual reconstructed module is proposed, which captures long-range dependencies in the network and extracts more critical contextual information, thus solving the problem of network degradation and accuracy degradation. Thirdly, an adaptive spatial feature fusion object detection head is constructed, which can directly learn how to filter different features at different feature layers spatially; useless information is filtered out, and only useful information is kept for combination to enhance the detection capability of the network further. Finally, network for underwater object detection is proposed based on the above three techniques. Extensive experiments were conducted on the well-known datasets of RUOD, Aquarium, URPC, and MS COCO. Compared to the prior state-of-the-art methods, the experimental findings demonstrate that the proposed approach obtains the highest mAP of 88.7%, 86.5%, 98.9%, and 71.4%, respectively. This represents an improvement of 1.2, 1.5, 8.5, and 0.2 percentage, in that order. The proposed model shows the capacity to function by applying self-attention to local details, as well as the capacity to grasp global long-range relationships, prioritize essential data, and spatially filter irrelevant information.

Single Shot Feature Aggregation Network For Underwater Object Detection

An Automatic Detection and Statistical Method for Underwater Fish Based on Foreground Region Convolution Network (FR-CNN)

Multi-scale aggregation feature pyramid with cornerness for underwater object detection

Research on Underwater Small Target Detection Algorithm Based on Improved YOLOv3

A Robust Underwater Multiclass Fish-School Tracking Algorithm

A Novel Underwater Image Enhancement and Improved Underwater Biological Detection Pipeline

Single-Stage Underwater Target Detection Based on Feature Anchor Frame Double Optimization Network

Multi-Scale Feature Fusion Enhancement for Underwater Object Detection

YOLO‐UOD: An underwater small object detector via improved efficient layer aggregation network

Self-attention and long-range relationship capture network for underwater object detection

Underwater small and occlusion object detection with feature fusion and global context decoupling head-based YOLO

An Efficient Convolutional Network for Underwater Object Detection

Dual Spatial Attention Network for Underwater Object Detection With Sonar Imagery

Underwater Species Detection using Channel Sharpening Attention

Underwater Object Recognition Based on Deep Encoding-Decoding Network

Lightweight Underwater Object Detection Algorithm for Embedded Deployment Using Higher-Order Information and Image Enhancement

Underwater Target Detection Based on Parallel High-Resolution Networks

Feature-adaptive FPN with multiscale context integration for underwater object detection

Underwater Object Detection Algorithm Based on an Improved YOLOv8

Accelerating Fish Detection and Recognition by Sharing CNNs with Objectness Learning

UGC-YOLO: Underwater Environment Object Detection Based on YOLO with a Global Context Block