Maritime Perception with Sonar & Near-Infrared

Problem

Autonomous underwater and surface vehicles must detect objects where cameras struggle (murky water, low light and glare), and labeled sonar and near-infrared data is scarce.

Approach

• Built sonar and near-infrared datasets from the ground up
• Sensor-specific preprocessing, annotation and augmentation pipelines
• Trained and evaluated object-detection models for complex maritime scenes

Result

Two papers at IEEE i-COSTE 2023: one on sonar-based detection for AUVs and one on near-infrared detection for ASVs.