Multiple human tracking is a fundamental problem in understanding the context of a visual scene. Although both accuracy and speed are required in real-world applications, recent tracking methods based on deep learning focus on accuracy and require a substantial amount of running time. We aim to improve tracking running speeds by performing human detections at certain frame intervals because it accounts for most of the running time. The question is how to maintain accuracy while skipping human detection. In this paper, we propose a method that interpolates the detection results by using an optical flow, which is based on the fact that someone's appearance does not change much between adjacent frames. To maintain the tracking accuracy, we introduce robust interest point detection within the human regions and a tracking termination metric defined by the distribution of the interest points. On the MOT17 and MOT20 datasets in the MOTChallenge, the proposed SDOF-Tracker achieved the best performance in terms of total running time while maintaining the MOTA metric. Our code is available at https://github.com/hitottiez/sdof-tracker.
Hitoshi NISHIMURA
KDDI Research, Inc.
Satoshi KOMORITA
KDDI Research, Inc.
Yasutomo KAWANISHI
KDDI Research, Inc.,RIKEN,Nagoya University
Hiroshi MURASE
KDDI Research, Inc.,Nagoya University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Hitoshi NISHIMURA, Satoshi KOMORITA, Yasutomo KAWANISHI, Hiroshi MURASE, "SDOF-Tracker: Fast and Accurate Multiple Human Tracking by Skipped-Detection and Optical-Flow" in IEICE TRANSACTIONS on Information,
vol. E105-D, no. 11, pp. 1938-1946, November 2022, doi: 10.1587/transinf.2022EDP7022.
Abstract: Multiple human tracking is a fundamental problem in understanding the context of a visual scene. Although both accuracy and speed are required in real-world applications, recent tracking methods based on deep learning focus on accuracy and require a substantial amount of running time. We aim to improve tracking running speeds by performing human detections at certain frame intervals because it accounts for most of the running time. The question is how to maintain accuracy while skipping human detection. In this paper, we propose a method that interpolates the detection results by using an optical flow, which is based on the fact that someone's appearance does not change much between adjacent frames. To maintain the tracking accuracy, we introduce robust interest point detection within the human regions and a tracking termination metric defined by the distribution of the interest points. On the MOT17 and MOT20 datasets in the MOTChallenge, the proposed SDOF-Tracker achieved the best performance in terms of total running time while maintaining the MOTA metric. Our code is available at https://github.com/hitottiez/sdof-tracker.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2022EDP7022/_p
Copy
@ARTICLE{e105-d_11_1938,
author={Hitoshi NISHIMURA, Satoshi KOMORITA, Yasutomo KAWANISHI, Hiroshi MURASE, },
journal={IEICE TRANSACTIONS on Information},
title={SDOF-Tracker: Fast and Accurate Multiple Human Tracking by Skipped-Detection and Optical-Flow},
year={2022},
volume={E105-D},
number={11},
pages={1938-1946},
abstract={Multiple human tracking is a fundamental problem in understanding the context of a visual scene. Although both accuracy and speed are required in real-world applications, recent tracking methods based on deep learning focus on accuracy and require a substantial amount of running time. We aim to improve tracking running speeds by performing human detections at certain frame intervals because it accounts for most of the running time. The question is how to maintain accuracy while skipping human detection. In this paper, we propose a method that interpolates the detection results by using an optical flow, which is based on the fact that someone's appearance does not change much between adjacent frames. To maintain the tracking accuracy, we introduce robust interest point detection within the human regions and a tracking termination metric defined by the distribution of the interest points. On the MOT17 and MOT20 datasets in the MOTChallenge, the proposed SDOF-Tracker achieved the best performance in terms of total running time while maintaining the MOTA metric. Our code is available at https://github.com/hitottiez/sdof-tracker.},
keywords={},
doi={10.1587/transinf.2022EDP7022},
ISSN={1745-1361},
month={November},}
Copy
TY - JOUR
TI - SDOF-Tracker: Fast and Accurate Multiple Human Tracking by Skipped-Detection and Optical-Flow
T2 - IEICE TRANSACTIONS on Information
SP - 1938
EP - 1946
AU - Hitoshi NISHIMURA
AU - Satoshi KOMORITA
AU - Yasutomo KAWANISHI
AU - Hiroshi MURASE
PY - 2022
DO - 10.1587/transinf.2022EDP7022
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E105-D
IS - 11
JA - IEICE TRANSACTIONS on Information
Y1 - November 2022
AB - Multiple human tracking is a fundamental problem in understanding the context of a visual scene. Although both accuracy and speed are required in real-world applications, recent tracking methods based on deep learning focus on accuracy and require a substantial amount of running time. We aim to improve tracking running speeds by performing human detections at certain frame intervals because it accounts for most of the running time. The question is how to maintain accuracy while skipping human detection. In this paper, we propose a method that interpolates the detection results by using an optical flow, which is based on the fact that someone's appearance does not change much between adjacent frames. To maintain the tracking accuracy, we introduce robust interest point detection within the human regions and a tracking termination metric defined by the distribution of the interest points. On the MOT17 and MOT20 datasets in the MOTChallenge, the proposed SDOF-Tracker achieved the best performance in terms of total running time while maintaining the MOTA metric. Our code is available at https://github.com/hitottiez/sdof-tracker.
ER -