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

3D object detection with multi-sensors is essential for an accurate andreliable perception system of autonomous driving and robotics. Existing 3Ddetectors significantly improve the accuracy by adopting a two-stage paradigmwhich merely relies on LiDAR point clouds for 3D proposal refinement. Thoughimpressive, the sparsity of point clouds, especially for the points far away,making it difficult for the LiDAR-only refinement module to accuratelyrecognize and locate objects.To address this problem, we propose a novelmulti-modality two-stage approach named FusionRCNN, which effectively andefficiently fuses point clouds and camera images in the Regions ofInterest(RoI). FusionRCNN adaptively integrates both sparse geometryinformation from LiDAR and dense texture information from camera in a unifiedattention mechanism. Specifically, it first utilizes RoIPooling to obtain animage set with a unified size and gets the point set by sampling raw pointswithin proposals in the RoI extraction step; then leverages an intra-modalityself-attention to enhance the domain-specific features, following by awell-designed cross-attention to fuse the information from twomodalities.FusionRCNN is fundamentally plug-and-play and supports differentone-stage methods with almost no architectural changes. Extensive experimentson KITTI and Waymo benchmarks demonstrate that our method significantly booststhe performances of popular detectors.Remarkably, FusionRCNN significantlyimproves the strong SECOND baseline by 6.14% mAP on Waymo, and outperformscompeting two-stage approaches. Code will be released soon athttps://github.com/xxlbigbrother/Fusion-RCNN.

FusionRCNN: LiDAR-Camera Fusion for Two-stage 3D Object Detection