LMOT: Efficient Light-Weight Detection and Tracking in Crowds
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...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2022-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9852199/ |
| _version_ | 1828299641391677440 |
|---|---|
| author | Rana Mostafa Hoda Baraka Abdelmoniem Bayoumi |
| author_facet | Rana Mostafa Hoda Baraka Abdelmoniem Bayoumi |
| author_sort | Rana Mostafa |
| collection | DOAJ |
| description | 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-the-art trackers while being only 3.8% behind in performance accuracy on average leading to a much computationally lighter model. |
| first_indexed | 2024-04-13T12:57:38Z |
| format | Article |
| id | doaj.art-55277569243046348e990e4d951b6083 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-13T12:57:38Z |
| publishDate | 2022-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-55277569243046348e990e4d951b60832022-12-22T02:46:00ZengIEEEIEEE Access2169-35362022-01-0110830858309510.1109/ACCESS.2022.31971579852199LMOT: Efficient Light-Weight Detection and Tracking in CrowdsRana Mostafa0https://orcid.org/0000-0001-9097-1449Hoda Baraka1Abdelmoniem Bayoumi2https://orcid.org/0000-0002-1334-8095Department of Computer Engineering, Faculty of Engineering, Cairo University, Giza, EgyptDepartment of Computer Engineering, Faculty of Engineering, Cairo University, Giza, EgyptDepartment of Computer Engineering, Faculty of Engineering, Cairo University, Giza, EgyptMulti-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-the-art trackers while being only 3.8% behind in performance accuracy on average leading to a much computationally lighter model.https://ieeexplore.ieee.org/document/9852199/Multi-object trackingpedestrian trackingjoint detection and trackingobject detectiondeep learning |
| spellingShingle | Rana Mostafa Hoda Baraka Abdelmoniem Bayoumi LMOT: Efficient Light-Weight Detection and Tracking in Crowds IEEE Access Multi-object tracking pedestrian tracking joint detection and tracking object detection deep learning |
| title | LMOT: Efficient Light-Weight Detection and Tracking in Crowds |
| title_full | LMOT: Efficient Light-Weight Detection and Tracking in Crowds |
| title_fullStr | LMOT: Efficient Light-Weight Detection and Tracking in Crowds |
| title_full_unstemmed | LMOT: Efficient Light-Weight Detection and Tracking in Crowds |
| title_short | LMOT: Efficient Light-Weight Detection and Tracking in Crowds |
| title_sort | lmot efficient light weight detection and tracking in crowds |
| topic | Multi-object tracking pedestrian tracking joint detection and tracking object detection deep learning |
| url | https://ieeexplore.ieee.org/document/9852199/ |
| work_keys_str_mv | AT ranamostafa lmotefficientlightweightdetectionandtrackingincrowds AT hodabaraka lmotefficientlightweightdetectionandtrackingincrowds AT abdelmoniembayoumi lmotefficientlightweightdetectionandtrackingincrowds |