FR-DETR: Frequency and Recurrent Feature Refinement for Robust Object Detection under Adverse Weather
2026-06-29 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors address the problem of detecting objects in bad weather, which is hard because images get blurry or noisy. Instead of improving the entire image first (which is slow and not very effective), they focus on improving important parts of the image using special tricks with image frequencies. They created two modules: one that adjusts low and high frequency details to better tell objects from background, and another that keeps refining the important features step-by-step. Their method, FR-DETR, works better and faster than previous methods that enhanced whole images before detection.
object detectionadverse weatherfrequency domainfeature refinementforeground-background discriminationfrequency refinement modulerecurrent focus refinement moduledomain shiftimage enhancementcomputational efficiency
Authors
Tuan-Duc Nguyen, Duc-Trong Le
Abstract
Object detection under adverse weather remains challenging due to severe visual degradations and domain shifts. Existing enhancer-based approaches attempt to improve detection by cascading an enhancer with a detector, but they introduce redundant feature extraction and incur high computational cost with limited accuracy gains when paired with SOTA detectors. We propose FR-DETR, a detector-centric framework that refines features rather than images, focusing enhancement on regions of interest and leveraging frequency-domain cues. Specifically, we design (I) a Frequency Refinement Module that dynamically separates and reweights low- and high-frequency components to improve foreground-background discrimination, and (II) a Recurrent Focus Refinement Module (RFRM) that iteratively refines features using coarse predictions as guidance. Extensive experiments demonstrate that FR-DETR achieves superior detection accuracy under adverse weather while being significantly more computationally efficient than enhancer-based methods. Our implementation is available at https://github.com/ducnt1210/FR-DETR.