End-to-end Learning of Multi-sensor 3D Tracking by Detection
Davi Frossard, Raquel Urtasun
2018-06-29Multiple Object Tracking
Abstract
In this paper we propose a novel approach to tracking by detection that can exploit both cameras as well as LIDAR data to produce very accurate 3D trajectories. Towards this goal, we formulate the problem as a linear program that can be solved exactly, and learn convolutional networks for detection as well as matching in an end-to-end manner. We evaluate our model in the challenging KITTI dataset and show very competitive results.
Results
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Video | KITTI Test (Online Methods) | MOTA | 76.15 | DSM |
| Object Tracking | KITTI Test (Online Methods) | MOTA | 76.15 | DSM |
| Multiple Object Tracking | KITTI Test (Online Methods) | MOTA | 76.15 | DSM |
Related Papers
Learning better representations for crowded pedestrians in offboard LiDAR-camera 3D tracking-by-detection2025-05-21LiDAR MOT-DETR: A LiDAR-based Two-Stage Transformer for 3D Multiple Object Tracking2025-05-19Using Cross-Domain Detection Loss to Infer Multi-Scale Information for Improved Tiny Head Tracking2025-05-14Towards Accurate State Estimation: Kalman Filter Incorporating Motion Dynamics for 3D Multi-Object Tracking2025-05-12History-Aware Transformation of ReID Features for Multiple Object Tracking2025-03-16OVTR: End-to-End Open-Vocabulary Multiple Object Tracking with Transformer2025-03-13INTACT: Inducing Noise Tolerance through Adversarial Curriculum Training for LiDAR-based Safety-Critical Perception and Autonomy2025-02-04Heterogeneous Graph Transformer for Multiple Tiny Object Tracking in RGB-T Videos2024-12-14