# Underwater Object Detection

Underwater cultural heritage detection has lacked dedicated annotated data, and RGB imagery suffers color shift and weakened boundaries from selective absorption and scattering. The new WRF dataset gathers 4,732 images from the Gulf of Mexico, the Black Sea and the Mediterranean, with 17,059 annotated instances across 21 artifact and ship-component categories plus depth maps estimated by Depth Anything V2. UCH-Net extracts appearance and geometry through dual RGB and depth streams and modulates color features by local color bias, reaching 77.5% mAP and 88.0% mAP75 on WRF and 66.0% mAP on RUOD.

Underwater cultural heritage detection has lacked dedicated annotated data, and RGB imagery suffers color shift and weakened boundaries from selective absorption and scattering. The new WRF dataset gathers 4,732 images from the Gulf of Mexico, the Black Sea and the Mediterranean, with 17,059 annotated instances across 21 artifact and ship-component categories plus depth maps estimated by Depth Anything V2. UCH-Net extracts appearance and geometry through dual RGB and depth streams and modulates color features by local color bias, reaching 77.5% mAP and 88.0% mAP75 on WRF and 66.0% mAP on RUOD.

Underwater images suffer color shifts, low contrast and blurred edges from light attenuation, particle scattering and uneven illumination, while public datasets cover only a limited range of degradation types and intensities. GDH-Det extracts water color, turbidity and blur level from real images as degradation conditions, uses a conditional diffusion network with bounding-box masks to generate degraded images at continuously varying intensities, and screens candidates by object-region structure and detection credibility. During detector training, sample difficulty is computed from classification uncertainty, localization deviation and perturbation response to set sampling probabilities, reaching mAP@0.5 of 85.3%, 79.6%, 66.2% and 84.7% on URPC2020, DUO, UODD and Brackish.

Underwater images suffer color shifts, low contrast and blurred edges from light attenuation, particle scattering and uneven illumination, while public datasets cover only a limited range of degradation types and intensities. GDH-Det extracts water color, turbidity and blur level from real images as degradation conditions, uses a conditional diffusion network with bounding-box masks to generate degraded images at continuously varying intensities, and screens candidates by object-region structure and detection credibility. During detector training, sample difficulty is computed from classification uncertainty, localization deviation and perturbation response to set sampling probabilities, reaching mAP@0.5 of 85.3%, 79.6%, 66.2% and 84.7% on URPC2020, DUO, UODD and Brackish.