SAGA-KF
Sensor-Agnostic Graph-Aware Kalman Filter for Multi-Modal Multi-Object Tracking
Depanshu Sani, Anirudh R Iyer, Prakhar Rai, Saket Anand, Anuj Srivastava, and Kaushik Kalyanaraman
 (Accepted at ICPR 2024)
Depanshu Sani, Anirudh R Iyer, Prakhar Rai, Saket Anand, Anuj Srivastava, and Kaushik Kalyanaraman
 (Accepted at ICPR 2024)
Recent progress in open-source object detection techniques has significantly advanced Multi-Object Tracking (MOT) methodologies, primarily under the tracking-by-detection paradigm. To enhance the robustness and reliability of MOT systems, recent research has proposed integrating information gathered from diverse sensors. However, many Kalman filter-based MOT approaches assume the independence of object trajectories, overlooking potential inter-object relationships. While some efforts have been made to incorporate these relationships, they often concentrate on learning feature representations to facilitate better association. Moreover, the existing filter-based method for estimating graphs from noisy data is unsuitable for online MOT applications. To alleviate these problems, we introduce a Sensor Agnostic Graph-Aware (SAGA) Kalman filter, which is the first online state estimation technique designed to fuse multi-modal graphs derived from noisy multi-sensor data. We validate the effectiveness of our proposed framework through extensive experiments conducted on both synthetic and real-world driving dataset (nuScenes). Our results showcase an improvement in MOTA and a reduction in estimated position errors (MOTP) and identity switches (IDS) for tracked objects using the SAGA-KF.
This work was supported by the NSF-TIH grant <TIH/ iHub Drishti/Project/2022-23/22>. We are grateful for the support from the TIH, iHub Drishti at IIT Jodhpur and the Infosys Center for Artificial Intelligence at IIIT-Delhi. We also appreciate the contributions of Aditi Basu Bal and Nilay Shrivastava during the early stages of the project.
@InProceedings{10.1007/978-3-031-78444-6_25,
author="Sani, Depanshu
and Iyer, Anirudh
and Rai, Prakhar
and Anand, Saket
and Srivastava, Anuj
and Kalyanaraman, Kaushik",
editor="Antonacopoulos, Apostolos
and Chaudhuri, Subhasis
and Chellappa, Rama
and Liu, Cheng-Lin
and Bhattacharya, Saumik
and Pal, Umapada",
title="Sensor-Agnostic Graph-Aware Kalman Filter for Multi-Modal Multi-Object Tracking",
booktitle="Pattern Recognition",
year="2025",
publisher="Springer Nature Switzerland",
address="Cham",
pages="380--398",
abstract="Recent progress in open-source object detection techniques has significantly advanced Multi-Object Tracking (MOT) methodologies, primarily under the tracking-by-detection paradigm. To enhance the robustness and reliability of MOT systems, recent research has proposed integrating information gathered from diverse sensors. However, many Kalman filter-based MOT approaches assume the independence of object trajectories, overlooking potential inter-object relationships. While some efforts have been made to incorporate these relationships, they often concentrate on learning feature representations to facilitate better association. Moreover, the existing filter-based method for estimating graphs from noisy data is unsuitable for online MOT applications. To alleviate these problems, we introduce a Sensor Agnostic Graph-Aware (SAGA) Kalman filter, which is the first online state estimation technique designed to fuse multi-modal graphs derived from noisy multi-sensor data. We validate the effectiveness of our proposed framework through extensive experiments conducted on both synthetic and real-world driving dataset (nuScenes). Our results showcase an improvement in MOTA and a reduction in estimated position errors (MOTP) and identity switches (IDS) for tracked objects using the SAGA-KF.",
isbn="978-3-031-78444-6"
}
For any queries, reach out to us at depanshus@iiitd.ac.in