YOLO-Underwater: A Real-Time Object Detection Framework for Enhanced Underwater Robotics Operations

We propose a YOLOv7-underwater model for real-time underwater object detection, specifically designed to meet the requirements of underwater robotics. The model integrates a new ConvNeXt convolutional layer structure and a wide receptive field module, incorporating techniques such as inverted bottleneck layers, GELU activation functions, and layer normalization. Additionally, it introduces a parameter-free attention module (SimAM) to enhance network performance, addressing challenges posed by varying water conditions and image blurriness. Experimental results demonstrate that the proposed model significantly improves the efficiency and accuracy of underwater object detection and recognition compared to other algorithms, making it suitable for real-time applications in diverse underwater environments.