The KITTI Vision Benchmark Suite

BindTrack is a multimodal 3D multi-object tracking
framework that maintains persistent cross-modal
bindings between 3D and 2D trajectories. These
bindings provide cross-modal existence priors for
identifying temporarily missed 3D objects. A
lightweight recovery module further estimates
missing 3D states using trajectory history, point-
cloud observations, and optional 2D guidance. The
method improves tracking robustness under occlusion
and sparse LiDAR observations.

input_score_3d=0.30, offline=true

@inproceedings{wu2026bindtrack,
title={BindTrack: Binding Cross-Modal
Trajectories for Robust 3D Multi-Object Tracking},
author={Weidong Wu and others},
booktitle={Asian Conference on Computer Vision
(ACCV)},
year={2026},
note={To appear}
}