Pedestrian detector boosts accuracy with contour and spectral cues
Contour-Guided Spectral Routing for Robust Real-Time Pedestrian Detection
Computer Vision and Pattern Recognition
Summary
Detecting pedestrians quickly and reliably in all kinds of weather and lighting is hard, especially when people are far away or partly hidden. The authors propose a new way to process images that carefully mixes boundary shapes, frequency-based information, and attention to focus on important details. Their method treats these parts as one system that first gathers broad context, then combines low- and high-frequency signals, and finally checks disagreements near edges to improve detection. Tested on a standard driving scene dataset, their approach improved accuracy compared to previous real-time methods, especially in challenging conditions like fog or low light.
What this means in practice
- •For automotive perception teams: Improve real-time pedestrian detection in self-driving cars under difficult weather and lighting using contour-guided spectral routing.$Commercial implications: Enables safer pedestrian detection products for autonomous vehicle manufacturers by improving accuracy and robustness in challenging conditions.
- •For urban traffic monitoring teams: Enhance pedestrian counting and monitoring in city cameras by incorporating boundary and frequency based features to reduce missed detections.
Authors
Sam Williams, Yuan Xiang
Abstract
Real-time pedestrian detection in driving scenes is constrained by three coupled failure modes: tiny targets lose discriminative evidence, occlusion weakens geometric support, and weather or illumination changes distort appearance statistics. We formulate the detector through a unified \emph{contour-guided spectral routing} view rather than treating frequency processing, attention, and boundary reasoning as independent add-ons. The detector routes information in a prescribed order: spatial evidence is first augmented with global spectral context, deep representations then exchange spatial and spectral cues, and cross-scale fusion is finally conditioned on boundary--semantic disagreement. This ordering yields a compact representation pipeline in which low-frequency context stabilizes global structure while high-frequency evidence protects small-object contours. We further retain a wavelet-subband training transformation that perturbs low- and high-frequency coefficients independently, targeting appearance shifts caused by fog, rain, snow, and low illumination. The formulation exposes a single routing variable at each stage and distinguishes reusable signal transforms from the task-specific policy that decides where each signal is injected. On CityPersons, the proposed detector obtains 70.4 AP$_{50}$ and 44.2 AP$_{50:95}$, compared with 68.1 and 42.2 for RT-DETR, while the full wavelet-augmented configuration reaches 71.1 and 44.6.