A practical approach to underwater depth and surface normals estimation

Monocular Depth and Surface Normals Estimation (MDSNE) is crucial for tasks such as three-dimensional (3D) reconstruction, autonomous navigation, and underwater exploration. Current methods rely either on discriminative models, which struggle with transparent or reflective surfaces, or generative models, which, while accurate, are computationally expensive. This paper presents a novel deep learning model for Monocular Depth and Surface Normals Estimation (MDSNE), specifically tailored for underwater environments, using a hybrid architecture that integrates Convolutional Neural Networks (CNNs) with Transformers, leveraging the strengths of both approaches. Training effective MDSNE models is often hampered by noisy real-world datasets and the limited generalization of synthetic datasets. To address this, we generate pseudo-labeled real data using multiple pre-trained MDSNE models. To curate this data without ground-truth supervision, we propose the Depth Normal Evaluation and Selection Algorithm (DNESA), which scores each candidate teacher’s prediction with three intrinsic image-quality cues (multi-scale gradient alignment with the RGB image, image–depth mutual information, and edge-aware total variation) and selects the top-scoring pseudo-label per image. A lightweight student model is then trained on this curated dataset. AquaMDS uses only 7.58 million parameters, 44 × to 113 × smaller than the four state-of-the-art foundation-model teachers it distils from (DepthAnything V2, ZoeDepth, Metric3D V2, and Marigold), and runs at 63 FPS on a single GPU. Despite this efficiency, on USOD10K and Atlantis AquaMDS substantially outperforms every prior similarly-sized baseline method and is competitive with the giant teachers under a unified evaluation protocol; on the small Sea-Thru D3 split AquaMDS sits mid-pack among the giant teachers. Key contributions include: a novel and efficient MDSNE model, the DNESA algorithm, a domain-specific data pipeline, and a focus on real-time performance and scalability. Designed for real-world underwater applications, our model facilitates low-cost deployments in underwater robots and autonomous vehicles, bridging the gap between research and practical implementation.