BenthicDINO: Physics-Informed Self-Distillation for View-Invariant Side-Scan Sonar Representations
2026-08-24 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionArtificial IntelligenceMachine Learning
AI summaryⓘ
The authors tackled the problem of noisy and complex images from side-scan sonar by creating a new computer learning method that understands the physical causes of image distortions. They designed special image changes and mathematical rules to make the computer focus on the real seabed features instead of noise from how the image was taken. Their method groups seabed types accurately without needing labeled examples and works well even with very few labels. When tested, their approach showed strong accuracy in recognizing seabed patterns using much less annotated data than usual.
side-scan sonarself-supervised learningDINOv3ConvNeXt-v2speckle noiseHilbert-Schmidt Independence Criterion (HSIC)view-invariancesemantic segmentationmean Intersection over Union (mIoU)data efficiency
Authors
Taqi Hamoda, Hayat Rajani, Nuno Gracias
Abstract
Automated perception in side-scan sonar (SSS) imagery is severely hindered by physical acoustic artifacts, resulting in representations that inextricably mix intrinsic seabed reflectivity with transient viewing geometries. Existing self-supervised learning (SSL) frameworks rely on augmentations designed for natural images, failing to account for acoustic degradation and explicitly enforce view-invariance. To address this gap, we introduce a physics-informed self-distillation framework built upon the DINOv3 architecture utilizing a ConvNeXt-v2-Tiny backbone to maximize data efficiency. The proposed methodology enforces view-invariance through two primary mechanisms: physically motivated augmentations that simulate speckle noise, range-dependent attenuation, and radiometric miscalibration; and a Hilbert-Schmidt Independence Criterion (HSIC) penalty that explicitly decouples learned dense patch features from physical viewing parameters. Furthermore, we propose a dense, hierarchical feature fusion strategy across all four network stages to preserve fine-grained sediment details alongside deep semantic abstractions. Extensive evaluation demonstrates that the framework natively groups complex benthic topographies into stable, noise-free semantic clusters without relying on manual annotations. During supervised downstream tasks on the S3Seg dataset, the fused representations exhibited exceptional data efficiency, achieving 96% of its absolute peak performance using only 10% of the available annotated data, ultimately reaching a mean Intersection over Union (mIoU) of 71.4% and an overall accuracy of 86.5%.