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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

3D multi-object tracking (MOT) is vital for many applications includingautonomous driving vehicles and service robots. With the commonly usedtracking-by-detection paradigm, 3D MOT has made important progress in recentyears. However, these methods only use the detection boxes of the current frameto obtain trajectory-box association results, which makes it impossible for thetracker to recover objects missed by the detector. In this paper, we presentTrajectoryFormer, a novel point-cloud-based 3D MOT framework. To recover themissed object by detector, we generates multiple trajectory hypotheses withhybrid candidate boxes, including temporally predicted boxes and current-framedetection boxes, for trajectory-box association. The predicted boxes canpropagate object's history trajectory information to the current frame and thusthe network can tolerate short-term miss detection of the tracked objects. Wecombine long-term object motion feature and short-term object appearancefeature to create per-hypothesis feature embedding, which reduces thecomputational overhead for spatial-temporal encoding. Additionally, weintroduce a Global-Local Interaction Module to conduct information interactionamong all hypotheses and models their spatial relations, leading to accurateestimation of hypotheses. Our TrajectoryFormer achieves state-of-the-artperformance on the Waymo 3D MOT benchmarks. Code is available athttps://github.com/poodarchu/EFG .

TrajectoryFormer: 3D Object Tracking Transformer with Predictive Trajectory Hypotheses