Adaptive enhancement improves underwater object retrieval accuracy
QMSR: Query-Conditioned Mask-wise Expert Routing for Robust Open-Vocabulary Underwater Object Retrieval
Robotics
Summary
Finding specific underwater objects in images can be hard because water distorts pictures in different ways. The authors created a smart system that chooses the best way to improve each image based on the search query and candidate objects. This method decides how much to enhance images or keep them raw to get better search results. Their system showed much better accuracy than always using the same enhancement method.
What this means in practice
- •For marine researchers: Improve retrieval of underwater objects from image databases by adaptively selecting image enhancement methods per query and candidate object.
- •For environmental monitoring teams: Enhance accuracy of underwater object detection tools by combining raw and enhanced images based on query-specific routing decisions.
Authors
Fuming Zhang, Dongyue Huang, Junjie Wen, Lihua Xie
Abstract
Open-vocabulary object retrieval remains challenging in complex underwater environments. Although underwater image enhancement (UIE) can improve visual quality, fixed UIE strategies may even underperform the Raw representation in retrieval, indicating that enhancement should not be applied as a uniform preprocessing step. To address this problem, we propose \textbf{QMSR}, a query-conditioned mask-wise expert routing framework for underwater open-vocabulary retrieval. Specifically, QMSR selects one pretrained UIE expert for each query--candidate pair and predicts a continuous Raw--Expert fusion strength, enabling adaptive enhancement while preserving useful Raw semantics. During training, a privileged ranking oracle provides expert-selection and fusion-strength supervision, while an annealed soft-routing relaxation facilitates optimization of the hard Top-1 routing policy. Experiments show that QMSR improves NDCG@10 by 17.6\% over an image--query shared router, while consistently outperforming fixed UIE strategies and remaining effective on held-out query categories. These results demonstrate the effectiveness of query-conditioned and candidate-specific enhancement routing for underwater open-vocabulary retrieval.