# Improve Underwater Object Detection through YOLOv12 Architecture and Physics-informed Augmentation

arXiv: https://arxiv.org/abs/2506.23505 · HTML: https://mlsanity.com/paper/2506.23505/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 18 of its 46 references that are in the map.
- First posted on arXiv 30 Jun 2025.

## Builds on (most cited first, 18 of 18)

- [Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks](https://mlsanity.com/paper/1506.01497/) (2015)
- [End-to-End Object Detection with Transformers](https://mlsanity.com/paper/2005.12872/) (2020)
- [You Only Look Once: Unified, Real-Time Object Detection](https://mlsanity.com/paper/1506.02640/) (2015)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [SSD: Single Shot MultiBox Detector](https://mlsanity.com/paper/1512.02325/) (2015)
- [Fast R-CNN](https://mlsanity.com/paper/1504.08083/) (2015)
- [Rich feature hierarchies for accurate object detection and semantic segmentation](https://mlsanity.com/paper/1311.2524/) (2013)
- [YOLOv12: Attention-Centric Real-Time Object Detectors](https://mlsanity.com/paper/2502.12524/) (2025)
- [Speed/accuracy trade-offs for modern convolutional object detectors](https://mlsanity.com/paper/1611.10012/) (2016)
- [The Computational Limits of Deep Learning](https://mlsanity.com/paper/2007.05558/) (2020)
- [FSSD: Feature Fusion Single Shot Multibox Detector](https://mlsanity.com/paper/1712.00960/) (2017)
- [A Dataset And Benchmark Of Underwater Object Detection For Robot Picking](https://mlsanity.com/paper/2106.05681/) (2021)
- [A New Dataset, Poisson GAN and AquaNet for Underwater Object Grabbing](https://mlsanity.com/paper/2003.01446/) (2020)
- [Underwater Object Detection in the Era of Artificial Intelligence: Current, Challenge, and Future](https://mlsanity.com/paper/2410.05577/) (2024)
- [SU-YOLO: Spiking Neural Network for Efficient Underwater Object Detection](https://mlsanity.com/paper/2503.24389/) (2025)
- [You Sense Only Once Beneath: Ultra-Light Real-Time Underwater Object Detection](https://mlsanity.com/paper/2504.15694/) (2025)
- [U-DECN: End-to-End Underwater Object Detection ConvNet with Improved DeNoising Training](https://mlsanity.com/paper/2408.05780/) (2024)
- [EPBC-YOLOv8: An efficient and accurate improved YOLOv8 underwater detector based on an attention mechanism](https://mlsanity.com/paper/2502.05788/) (2025)

## Built on by (most cited first, 1 of 1)

- [An Empirical Study on the Robustness of YOLO Models for Underwater Object Detection](https://mlsanity.com/paper/2509.17561/) (2025)

## Abstract (arXiv)

Underwater object detection is crucial for autonomous navigation, environmental monitoring, and marine exploration, but it is severely hampered by light attenuation, turbidity, and occlusion. Current methods balance accuracy and computational efficiency, but they have trouble deploying in real-time under low visibility conditions. Through the integration of physics-informed augmentation techniques with the YOLOv12 architecture, this study advances underwater detection. With Residual ELAN blocks to preserve structural features in turbid waters and Area Attention to maintain large receptive fields for occluded objects while reducing computational complexity. Underwater optical properties are addressed by domain-specific augmentations such as turbulence adaptive blurring, biologically grounded occlusion simulation, and spectral HSV transformations for color distortion. Extensive tests on four difficult datasets show state-of-the-art performance, with Brackish data registering 98.30% mAP at 142 FPS. YOLOv12 improves occlusion robustness by 18.9%, small-object recall by 22.4%, and detection precision by up to 7.94% compared to previous models. The crucial role of augmentation strategy is validated by ablation studies. This work offers a precise and effective solution for conservation and underwater robotics applications.