English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Perception is crucial in the realm of autonomous driving systems, wherebird's eye view (BEV)-based architectures have recently reachedstate-of-the-art performance. The desirability of self-supervisedrepresentation learning stems from the expensive and laborious process ofannotating 2D and 3D data. Although previous research has investigatedpretraining methods for both LiDAR and camera-based 3D object detection, aunified pretraining framework for multimodal BEV perception is missing. In thisstudy, we introduce CALICO, a novel framework that applies contrastiveobjectives to both LiDAR and camera backbones. Specifically, CALICOincorporates two stages: point-region contrast (PRC) and region-awaredistillation (RAD). PRC better balances the region- and scene-levelrepresentation learning on the LiDAR modality and offers significantperformance improvement compared to existing methods. RAD effectively achievescontrastive distillation on our self-trained teacher model. CALICO's efficacyis substantiated by extensive evaluations on 3D object detection and BEV mapsegmentation tasks, where it delivers significant performance improvements.Notably, CALICO outperforms the baseline method by 10.5% and 8.6% on NDS andmAP. Moreover, CALICO boosts the robustness of multimodal 3D object detectionagainst adversarial attacks and corruption. Additionally, our framework can betailored to different backbones and heads, positioning it as a promisingapproach for multimodal BEV perception.

CALICO: Self-Supervised Camera-LiDAR Contrastive Pre-training for BEV Perception