الفهرس 
Title : Efficient light-weight detection and tracking in crowds
SummaryOnly 14 pages are availabe for public view
 |  | from | 106 |  |  |
| | | from | 106 | | |
Abstract
Multi-object tracking is a vital component in various robotics and computer vision
applications. However, existing multi-object tracking techniques trade off computation
runtime for tracking accuracy leading to challenges in deploying such pipelines in real-time
applications. This paper introduces a novel real-time model, LMOT, i.e., Light-weight Multi-
Object Tracker, that performs joint pedestrian detection and tracking. LMOT introduces a
simplified DLA-34 encoder network to extract detection features for the current image that
are computationally efficient. Furthermore, we generate efficient tracking features using
a linear transformer for the prior image frame and its corresponding detection heatmap.
After that, LMOT fuses both detection and tracking feature maps in a multi-layer scheme
and performs a two-stage online data association relying on the Kalman filter to generate
tracklets. We evaluated our model on the challenging real-world MOT16/17/20 datasets,
showing LMOT significantly outperforms the state-of-the-art trackers concerning runtime
while maintaining high robustness. LMOT is approximately ten times faster than state-of-theart
trackers while being only 3.8% behind in performance accuracy on average leading to a much computationally lighter model.