Feature-Centric Underwater Vision Models
DOI:
https://doi.org/10.5753/reic.2026.8511Keywords:
Underwater Computer Vision, Feature Enhancement, Object Detection, Object Classification, Object Tracking, Image Restoration, Autonomous Underwater VehiclesAbstract
Underwater image enhancement (UIE) models are often optimized for human visual perception, which does not necessarily translate to improved performance in automated vision tasks. This paper describes a consolidated research journey consisting of the AquaFeat ecosystem: a series of plug-and-play modules designed to enhance hierarchical features for downstream robotics tasks. We present the transition from AquaFeat, a detection-focused module, to AquaFeat+, which introduces global-scale attention for classification and tracking, and finally C-Feat. C-Feat integrates these feature-centric backbones into a Compact Model framework, achieving a remarkable inference speed of 280 FPS while maintaining state-of-the-art performance in complex underwater environments.
Downloads
Referências
Dawkins, M., Prior, J., Lewis, B., et al. (2024). Fishtrack23: An ensemble underwater dataset for multi-object tracking. In WACV, pages 7167–7176.
Fu, Z., Lin, H., Yang, Y., Chai, S. b., Sun, L., Huang, Y., and Ding, X. (2022a). Unsupervised underwater image restoration: From a homology perspective. In AAAI, volume 36, pages 643–651. DOI: 10.1609/aaai.v36i1.19944.
Fu, Z., Wang, W., Huang, Y., Ding, X., and Ma, K.-K. (2022b). Uncertainty inspired underwater image enhancement. In ECCV, pages 465–482. Springer. DOI: 10.1007/978-3-031-19797-0_27.
Hashmi, K. A., Kallempudi, G., Stricker, D., and Afzal, M. Z. (2023). Featenhancer: Enhancing hierarchical features for object detection and beyond under low-light vision. In ICCV, pages 6725–6735. DOI: 10.48550/arXiv.2308.03594.
Hou, G., Li, N., Zhuang, P. b., Li, K., Sun, H., and Li, C. (2023). Non-uniform illumination underwater image restoration via illumination channel sparsity prior. IEEE Trans. Circuits Syst. Video Technol., 34(2):799–814. DOI: 10.1109/TCSVT.2023.3290363.
Islam, M. J., Edge, C., Xiao, Y., et al. (2020). Semantic Segmentation of Underwater Imagery: Dataset and Benchmark. In IROS. DOI: 10.48550/arXiv.2004.01241.
Jiang, J., Ye, T., Chen, S., Chen, E., Liu, Y., Jun, S., Bai, J., and Chai, W. (2023). Five a+ network: You only need 9k parameters for underwater image enhancement. In BMVC. DOI: 10.48550/arXiv.2305.08824.
Khan, M. R., Negi, A., Kulkarni, A., Phutke, S. S., Vipparthi, S. K., and Murala, S. (2025). Phaseformer: Phase-based attention mechanism for underwater image restoration and beyond. In WACV, pages 9618–9629. IEEE. DOI: 10.1109/WACV61041.2025.00931.
Li, C., Guo, C., Ren, W., Cong, R., Hou, J., Kwong, S., and Tao, D. (2020). An underwater image enhancement benchmark dataset and beyond. IEEE Trans. Image Process., 29:4376–4389. DOI: 10.1109/TIP.2019.2955241.
Liu, C., Shu, X., Pan, L., Shi, J., and Han, B. (2023). Multiscale underwater image enhancement in RGB and HSV color spaces. IEEE Trans. Instrum. Meas., 72:1–14. DOI: 10.1109/TIM.2023.3298395.
Liu, S., Fan, H., Lin, S., Wang, Q., Ding, N., and Tang, Y. (2022a). Adaptive learning attention network for underwater image enhancement. IEEE Robot. Autom. Lett., 7(2):5326–5333. DOI: 10.1109/LRA.2022.3156176.
Liu, Z., Mao, H., Wu, C.-Y., et al. (2022b). A convnet for the 2020s. In CVPR, pages 11976–11986. DOI: 10.1109/CVPR52688.2022.01167.
Mello, C. D., Moreira, B. U., de Oliveira Evald, P. J. D., Drews, P. J. L., and da Costa Botelho, S. S. (2022). Underwater enhancement based on a self-learning strategy and attention mechanism for high-intensity regions. Comput. Graph., 107:264–276. DOI: 10.1016/j.cag.2022.08.003.
Nathan, O. B., Levy, D., Treibitz, T., and Rosenbaum, D. (2024). Osmosis: Rgbd diffusion prior for underwater image restoration. In ECCV, pages 302–319. DOI: 10.1007/978-3-031-73033-7_17.
Peng, L., Zhu, C., and Bian, L. (2023). U-shape transformer for underwater image enhancement. IEEE Trans. Image Process.. DOI: 10.1109/TIP.2023.3276332.
Saleh, A., Sheaves, M., Jerry, D., and Rahimi Azghadi, M. (2022). Adaptive uncertainty distribution in deep learning for unsupervised underwater image enhancement. Available at SSRN 4742854. DOI: 10.2139/ssrn.4362438.
Schein, T. T., De Almeira, G. P., Brião, S. L., De Bem, R. A., De Oliveira, F. G., and Drews-Jr, P. L. (2024). Udbe: Unsupervised diffusion-based brightness enhancement in underwater images. In ICMLA, pages 664–670. IEEE. DOI: 10.1109/ICMLA61862.2024.00096.
Silva, E. C., Schein, T. T., Ramos, J. D. G., Oliveira, F. G., and Drews-Jr, P. L. J. (2026). C-feat: A compact feature-centric network shattering training and inference latency in underwater vision. ICPR.
Silva, E. C., Schein, T. T., Ramos, J. D. G., Silva, E. L., Brião, S. L., Oliveira, F. G., and Drews, P. L. J. (2025a). Aquafeat+: an underwater vision learning-based enhancement method for object detection, classification, and tracking. In ICAR, pages 432–437. DOI: 10.1109/ICAR65334.2025.11338647.
Silva, E. d. C., Schein, T. T., Brião, S. L., Costa, G. L. M., Oliveira, F. G., Almeida, G. P., Silva, E. L., Devincenzi, S. d. S., Machado, K. d. S., and Drews-Jr, P. L. J. (2025b). Aquafeat: A features-based image enhancement model for underwater object detection. In SIBGRAPI. DOI: 10.1109/SIBGRAPI67909.2025.11223382.
Ultralytics (2023). YOLO-v8. GitHub repository.
Wang, A., Chen, H., Liu, L., et al. (2024). Yolov10: Real-time end-to-end object detection. NeurIPS, 37:107984–108011. DOI: 10.48550/arXiv.2405.14458.
Xie, S., Girshick, R., Dollár, P., Tu, Z., and He, K. (2017). Aggregated residual transformations for deep neural networks. In CVPR. DOI: 10.48550/arXiv.1611.05431.
Zhou, J., He, Z., Lam, K.-M., Wang, Y., Zhang, W., Guo, C., and Li, C. (2024). AMSP-UOD: When vortex convolution and stochastic perturbation meet underwater object detection. In AAAI, volume 38. DOI: 10.48550/arXiv.2308.11918.
Zhuang, P., Wang, Y., and Qiao, Y. (2018). Wildfish: A large benchmark for fish recognition in the wild. In 26th ACM international conference on Multimedia, pages 1301–1309. DOI: 10.1145/3240508.324061.
Downloads
Published
Como Citar
Issue
Section
Licença
Copyright (c) 2026 Os autores
Este trabalho está licenciado sob uma licença Creative Commons Attribution 4.0 International License.
