Transformer-Based Multi-Object Smoothing with Decoupled Data Association and Smoothing
This work addresses computational bottlenecks in multi-object tracking for applications requiring accurate models with low-dimensional measurements, though it is incremental as it builds on existing DL approaches.
The paper tackles the multi-object smoothing problem in tracking by proposing a novel deep learning architecture that decouples data association from smoothing, addressing computational complexity issues in state-of-the-art methods. It provides the first comparison between traditional Bayesian and DL trackers in this setting, showing competitive performance across tasks of varying difficulty.
Multi-object tracking (MOT) is the task of estimating the state trajectories of an unknown and time-varying number of objects over a certain time window. Several algorithms have been proposed to tackle the multi-object smoothing task, where object detections can be conditioned on all the measurements in the time window. However, the best-performing methods suffer from intractable computational complexity and require approximations, performing suboptimally in complex settings. Deep learning based algorithms are a possible venue for tackling this issue but have not been applied extensively in settings where accurate multi-object models are available and measurements are low-dimensional. We propose a novel DL architecture specifically tailored for this setting that decouples the data association task from the smoothing task. We compare the performance of the proposed smoother to the state-of-the-art in different tasks of varying difficulty and provide, to the best of our knowledge, the first comparison between traditional Bayesian trackers and DL trackers in the smoothing problem setting.