Refine-and-Contrast: Adaptive Instance-Aware BEV Representations for Multi-UAV Collaborative Object Detection

TL;DR

AdaBEV introduces an adaptive, instance-aware BEV representation framework for multi-UAV 3D detection, improving accuracy and efficiency by refining foreground regions and enhancing feature discrimination.

Contribution

The paper proposes AdaBEV, a novel adaptive BEV framework with a refine-and-contrast paradigm, including BG-RM and IBCL modules, for improved multi-UAV collaborative detection.

Findings

Outperforms state-of-the-art methods at low resolutions

Achieves superior accuracy-computation trade-offs

Maintains low-resolution BEV inputs with negligible overhead

Abstract

Multi-UAV collaborative 3D detection enables accurate and robust perception by fusing multi-view observations from aerial platforms, offering significant advantages in coverage and occlusion handling, while posing new challenges for computation on resource-constrained UAV platforms. In this paper, we present AdaBEV, a novel framework that learns adaptive instance-aware BEV representations through a refine-and-contrast paradigm. Unlike existing methods that treat all BEV grids equally, AdaBEV introduces a Box-Guided Refinement Module (BG-RM) and an Instance-Background Contrastive Learning (IBCL) to enhance semantic awareness and feature discriminability. BG-RM refines only BEV grids associated with foreground instances using 2D supervision and spatial subdivision, while IBCL promotes stronger separation between foreground and background features via contrastive learning in BEV space. Extensive experiments on the Air-Co-Pred dataset demonstrate that AdaBEV achieves superior accuracy-computation trade-offs across model scales, outperforming other state-of-the-art methods at low resolutions and approaching upper bound performance while maintaining low-resolution BEV inputs and negligible overhead.