AquaProbe improves underwater vision model adaptation with targeted tuning
AquaProbe: Efficient Adaptation of Frozen Vision Models for Underwater Images
Computer Vision and Pattern RecognitionMachine LearningRobotics
Summary
Underwater images look very different from regular photos because water changes how light behaves. Vision models trained on normal pictures don’t work well underwater as a result. The authors created AquaProbe, a way to find where in a vision model the biggest differences happen when looking at underwater images. By using this information, they can adapt models more efficiently and improve underwater scene understanding without adding extra complexity.
What this means in practice
- •For autonomous underwater vehicle teams: Improve underwater segmentation accuracy by adapting pre-trained vision models with vehicle-specific water condition measurement.
- •For marine environmental monitoring teams: Enhance underwater image analysis for better habitat mapping by using AquaProbe to allocate adaptation resources efficiently.
Authors
Md Shihabul Islam Shovo, Nishi Kanta Paul, Trung Dong, Xiaomin Lin
Abstract
Reliable visual perception is essential for safe autonomous underwater exploration. Yet vision models pretrained on terrestrial images become less reliable underwater, where turbidity, backscatter, and wavelength-dependent attenuation alter scene appearance. Existing adaptation methods either spread adapters equally across the encoder depth or select layers using the divergence between two fixed domains. In this paper, we introduce AquaProbe for measuring how the location of the domain gap changes with water type and imaging range. AquaProbe separates newly introduced divergence from that accumulated through a frozen vision encoder using paired clean and physically rendered images and allocates a fixed LoRA budget according to the measured profile. The segmentation mean Intersection-over-Union (mIoU) for the proportional LoRA rank allocation, based on the measured profile, is 70.1; this is a better result than uniform LoRA rank allocation by 1.9 mIoU points while having the same parameter budget of 0.59 million adapter parameters. The three-way router, conditioned on estimated condition clusters, achieves 70.6 mIoU on the standard SUIM split and 63.4 AUDRC under held-out degradation.