Deep Fusion of Multi-Object Densities Using Transformer
Li, Lechi, Dai, Chen, Xia, Yuxuan, Svensson, Lennart
–arXiv.org Artificial Intelligence
In this paper, we demonstrate that deep learning based method can be used to fuse multi-object densities. Given a scenario with several sensors with possibly different field-of-views, tracking is performed locally in each sensor by a tracker, which produces random finite set multi-object densities. To fuse outputs from different trackers, we adapt a recently proposed transformer-based multi-object tracker, where the fusion result is a global multi-object density, describing the set of all alive objects at the current time. We compare the performance of the transformer-based fusion method with a well-performing model-based Bayesian fusion method in several simulated scenarios with different parameter settings using synthetic data. The simulation results show that the transformer-based fusion method outperforms the model-based Bayesian method in our experimental scenarios.
arXiv.org Artificial Intelligence
Feb-16-2023
- Country:
- Europe > Sweden
- Vaestra Goetaland > Gothenburg (0.04)
- North America > United States
- Massachusetts > Norfolk County > Norwood (0.04)
- Europe > Sweden
- Genre:
- Research Report (0.70)
- Technology: