*Result*: Dual Atrous and Attention–Enhanced DeepLabV3+ with Hybrid Feature Fusion for Underwater Object Detection.

*Underwater object detection poses unique challenges due to poor image visibility, light scattering, turbidity, and color distortion, which severely affect traditional deep learning performance. To overcome these limitations, we propose DAConv-DLV3+, a novel deep learning framework that enhances the DeepLabV3 + architecture with Dual Atrous Convolution, Dual Attention Mechanisms (spatial and channel), and a hybrid handcrafted feature fusion strategy using Completed Local Binary Pattern (CLBP) and Gray-Level Co-occurrence Matrix (GLCM). The model leverages a comprehensive preprocessing pipeline – including image resizing, white balancing, CLAHE contrast enhancement, dehazing, normalization, and noise reduction – to improve raw underwater image quality. Features extracted from CLBP-GLCM are integrated with deep semantic features at intermediate layers, enriching the model's representational power. Experimental evaluations conducted on four benchmark datasets (UOD, UOT32, LS-URTD, and UIEB) demonstrate that the proposed model achieves superior performance with an average accuracy of 98.39%, precision of 98.30%, F1-score of 98.28%, and MSE as low as 0.0064. The execution time is reduced to 2.10 s, outperforming existing models such as FasterNet-YOLOv7 and G-Net in both accuracy and efficiency. The results validate the effectiveness of DAConv-DLV3 + in underwater environments, establishing its suitability for marine automation, intelligent sensing systems, and embedded vision platforms, thereby offering a valuable contribution to the field of computer vision and intelligent systems in signal processing. The proposed DAConv-DLV3 + effectively addresses underwater challenges such as turbidity, color distortion, and blur through dual atrous convolution, dual attention mechanisms, and hybrid CLBP-GLCM feature fusion, achieving high detection accuracy while maintaining computational efficiency. [ABSTRACT FROM AUTHOR]*