Clean image guidance improves object detection on degraded images unevenly
Task-Sensitive Geometry of Representation Transfer for Object Detection under Image Degradation
Computer Vision and Pattern Recognition
Summary
Detecting objects in blurry or degraded images is harder, but using clear images to guide training can help. The authors studied how knowledge from clean images changes how a model recognizes objects in degraded images. They found that improvements are uneven—some objects benefit more than others—and discovered a mathematical way to describe these changes. While this approach helps fix some missed detections, turning this understanding into better overall training methods remains difficult.
What this means in practice
- •For computer vision engineers: Improve detection accuracy on degraded images by incorporating clean-image supervision and targeted local representation interventions.
- •For machine learning model trainers: Use task-sensitive geometry to analyze and refine representation transfer effectiveness when training models under image quality degradation.
Authors
Van Vung Pham
Abstract
Object detection under image degradation can benefit from clean-image supervision, but aggregate gains do not imply that transferred representation changes are uniformly useful. We study how clean task knowledge affects degraded-image representations and whether local responses to structured representation directions can be characterized geometrically. Using paired clean and Gaussian-degraded BDD100K images, we show that clean-teacher distillation improves observed detection accuracy while producing heterogeneous object-level transfer. We isolate a representation component complementary to direct clean-teacher alignment and map it into the distilled student space through an orthogonal bridge. Controlled interventions rescue 13.22% of objects lost under the distilled representation, versus 4.30% under norm-matched random perturbations, with very low harm on preserved objects. We introduce task-sensitive geometry, a gradient-derived channel-space geometry constructed from normalized detection-loss gradients. On a reserved cohort, mapped-complement orientation within this frozen geometry is positively associated with local intervention-response magnitude after controlling for intervention magnitude (partial Spearman $ρ$ = 0.242, 95% CI [0.108, 0.359]). The relationship eplicates on independent data ($ρ$ = 0.180) and with RT-DETR-L ($ρ$ = 0.227), but not for the direct clean-teacher residual family, and it weakens for large interventions. Routing rules and specialized distillation objectives based on these signals do not yield statistically reliable gains over CLEANKD. These results support a local, direction-family-dependent task-sensitive geometry while showing that converting such structure into improved global training remains an open problem.