Architecture Overview

VDXNet is a lightweight aerial vehicle detection model built for intelligent transportation and remote-sensing scenarios. Its design replaces heavier feature aggregation components with RxDF modules that combine spatial and depthwise information efficiently. LiteFPP supports multiscale contextual representation, while CRDown reduces spatial dimensions with lower computational cost. This architecture is shaped around the practical difficulty of detecting small vehicles with different orientations and background clutter, while keeping the model efficient enough for real-time aerial monitoring workflows.

Architecture Diagram