24 December 2025

DfW Algorithm – Release Notes

(Optional) Ground-Truth Validation
An validation step is executed when ground-truth data are available in the input directory (automatically skipped otherwise).
(Optional) DMD Visualization
Mode-wise wave propagation videos derived from DMD can be generated in .mp4 format to visualize spatial wave patterns, controlled via the flag_show_DMD switch.
Persistent Runtime Logging
All console outputs produced during execution are saved to a .log file for post-run inspection and reproducibility.

Process-level Parallelization Redesign
The depth inversion workflow has been redesigned for true process-level parallelism, enabling efficient mode-wise and point-wise estimation on multi-core systems via joblib. Core routines in modules have also been refactored to be stateless , enabling safe parallel execution without side effects or race conditions.
Shared-Memory–Based Data Exchange
Intermediate data are exchanged through shared memory with read-only access, reducing memory overhead and inter-process communication while ensuring safe parallel execution.

Nonlinear wave correction module added Introduces a dedicated module to account for wave nonlinearity, improving depth inversion accuracy, particularly in shallow water and energetic wave conditions. Reference: Kim et al., 2026
Adaptive interrogation windowing implemented Automatically adjusts the optimal size and shape of the interrogation window with respect to global bottom slope, and local wavelength and direction.
Reference: Kim et al., 2025
Input and configuration handling refined
The input interface has been streamlined by supporting standard video formats (e.g., *.mov, *.mp4) instead of compressed NumPy archives. Spatial referencing is now decoupled from the video and provided via an external extent.txt file, while all runtime parameters have been consolidated into a standalone config.yaml file, removing hardcoded dependencies.
Visualization and reference data support extended
Visualization capabilities have been enhanced with optional satellite image overlays via basemap.png. In addition, ground truth inputs now support both raster-based formats (e.g., GeoTIFF) and point-based formats (e.g., *.csv, *.xyz), enabling more flexible validation and comparison workflows.

tags: