Traditional in-situ methods face challenges such as high deployment costs, poor spatial coverage, and susceptibility to storm conditions. This paper presents a video-based and HPC-enabled deep learning framework for sensor-free estimation of five coastal wave parameters: significant wave height (Hs), maximum wave height (Hmax), peak period (Tp), zero upcrossing period (Tz), and wave direction (theta).
The proposed architecture features a self-supervised V-JEPA ViT Small backbone for robust spatiotemporal feature extraction in visually challenging scenarios, a dual-stream SlowFast temporal encoder for broad bandwidth representation of wave motion in both hydrodynamic breaking and swell regimes, an optical flow stream based on the Farneback optical flow algorithm to add saliency information focusing on hydrodynamically active wavelength bands, and a multi-task regression layer with dispersion constraints (Airy wave dispersion \lambda_p = 0.1).
The model was trained on an NVIDIA DGX A100 cluster, early stopping at epoch 31, achieving Pearson correlation coefficients of 0.451, 0.578, 0.643, 0.680, and 0.832 for Hs, Hmax, Tp, Tz, and wave direction, respectively, demonstrating generalization ability to geographically diverse held-out test data sites. Operating in a data-limited regime (6 annotated training scenes), the framework shows statistically significant temporal correlations (PCC from 0.451 to 0.832), confirming proof of concept feasibility; R2 values (max 0.246) indicate that variance capture will improve with larger annotated datasets.
Blogger's Review: This study highlights the potential of combining HPC with deep learning, particularly in ocean monitoring where traditional methods fall short. The framework not only enhances the accuracy of wave parameter estimation but also provides new perspectives and tools for future marine research.