The Eighth Visual Object Tracking VOT2020 Challenge Results

Kristan, Matej; Leonardis, Aleš; Matas, Jiřı́; Felsberg, Michael; Pflugfelder, Roman; Kämäräinen, Joni-Kristian; Danelljan, Martin; Zajc, Luka Čehovin; Lukežič, Alan; Drbohlav, Ondřej; He, Linbo; Zhang, Yushan; Yan, Song; Yang, Jinyu; Fernández, Gustavo J.; Hauptmann, Alexander G.; Memarmoghadam, Alireza; García-Martín, Álvaro; Robinson, Andreas; Varfolomieiev, Anton; Gebrehiwot, Awet Haileslassie; Uzun, Bedirhan; Yan, Bin; Li, Bing; Chen, Qian; Tsai, Chi-Yi; Micheloni, Christian; Wang, Dong; Wang, Fei; Xie, Fei; Lawin, Felix Järemo; Gustafsson, Fredrik; Foresti, Gian Luca; Bhat, Goutam; Chen, Guangqi; Ling, Haibin; Zhang, Haitao; Çevikalp, Hakan; Zhao, Haojie; Bai, Haoran; Kuchibhotla, Hari Chandana; Sarıbaş, Hasan; Fan, Heng; Ghanei-Yakhdan, Hossein; Li, Houqiang; Peng, Houwen; Lu, Huchuan; Li, Hui; Khaghani, Javad; Bescós, Jesús; Li, Jianhua; Fu, Jianlong; Yu, Jiaqian; Xu, Jingtao; Kittler, Josef; Yin, Jianqin; Lee, Junhyun; Yu, Kaicheng; Liu, Kaiwen; Yang, Kang; Dai, Kenan; Cheng, Li; Zhang, Li; Wang, Lijun; Wang, Linyuan; Gool, Luc Van; Bertinetto, Luca; Dunnhofer, Matteo; Cheng, Miao; Dasari, Mohana Murali; Wang, Ning; Zhang, Pengyu; Torr, Philip H. S.; Wang, Qiang; Timofte, Radu; Gorthi, Rama Krishna Sai Subrahmanyam; Choi, Seokeon; Marvasti-Zadeh, Seyed Mojtaba; Zhao, Shaochuan; Kasaei, Shohreh; Qiu, Shoumeng; Chen, Shuhao; Schön, Thomas B.; Xu, Tianyang; Lü, Wei; Hu, Weiming; Zhou, Wengang; Qiu, Xi; Xiao, Ke; Wu, Xiao-Jun; Zhang, Xiaolin; Yang, Xiaoyun; Zhu, Xuefeng; Jiang, Yingjie; Wang, Ying‐Ming; Chen, Yiwei; Yu, Ye; Li, Yuezhou; Yao, Yuncon; Lee, Yunsung; Gu, Yuzhang; Wang, Zezhou; Tang, Zhangyong; Feng, Zhen-Hua; Mai, Zhijun; Zhang, Zhipeng; Wu, Zhirong; Ma, Zhimin

doi:10.1007/978-3-030-68238-5_39

Cited by 146 publications

(149 citation statements)

References 70 publications

Supporting

Mentioning

147

Contrasting

Order By: Relevance

“…Then, the refined candidate bounding box is sent into the localization module to determine the current status of the target, whether a risk of out-of-view exists and whether it is a reasonable position for target. Only reasonable candidate can be further checked by the verifier, where the feature distances, d (1) and d (2) , between the candidate object and objects of two feature pools are calculated. If any of the two distances is greater than the threshold, we use the Global Tracker [12] to find the most likely candidate boxB t (global) in constrained scope, which is determined by status of localization module.…”

Section: Methodsmentioning

confidence: 99%

“…For every candidate bounding boxB t , we simply use Euclidean method to calculate the distance d (1) betweenB t 's feature X (Bt) and the latest appearance feature pool X (1) as follows:…”

Section: Feature Calculation With Feature Poolmentioning

confidence: 99%

“…If d (1) and d (2) are smaller than the predefined thresholds θ 1 and θ 2 , respectively, then the candidateB t passes the verification.…”

Section: Feature Calculation With Feature Poolmentioning

confidence: 99%

“…Large-scale long-term tracking, whose length of benchmark (e.g., VOT2020LT [1] and LaSOT [2]) sequences is always at least the minute level, has attracted increasing attention due to its practicability of application. Compared with short-term tracking tasks, long-term tracking tasks have many additional challenging problems.…”

Section: Introductionmentioning

confidence: 99%

“…The top three performers in VOT2020LT challenge [1] adopted similar combination of three basic parts, namely, a short-term tracker for initial target localization, a global tracker to reidentify absent target, and an online trained verifier to distinguish whether the candidate target proposed by short-term tracker or global tracker is truly the target. Although the tracking framework is currently the most reason-able, the use of verifier and global tracker in actual tracking still has some problems.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A Method of Stable Long-Term Single Object Tracking

Tong

Zhao

et al. 2021

2021 IEEE International Conference on Multimedia and Expo (ICME)

View full text Add to dashboard Cite

We propose a stable long-term tracking method to deal with visual tracking in multiple complex scenarios to solve the problem of frequent disappearance and reappearance of targets in long-term tracking. In our method, we do not blindly start the global tracker once the target disappears, but use it only when necessary and in reasonable scope with the assistance of localization module. In addition, we designed an FP-verifier based on feature pools to reevaluate the candidate bounding boxes given by our local tracker and global tracker to ensure that online learning local tracker can be updated stably. Our method outperforms the state-of-the-art results on the VOT2020LT challenge. In addition, the control experiments show that our FP-verifier is more effective than the RT-MDNet verifier used by the top three winners of VOT2020LT challenge.

show abstract

Section: Methodsmentioning

confidence: 99%

“…For every candidate bounding boxB t , we simply use Euclidean method to calculate the distance d (1) betweenB t 's feature X (Bt) and the latest appearance feature pool X (1) as follows:…”

Section: Feature Calculation With Feature Poolmentioning

confidence: 99%

“…If d (1) and d (2) are smaller than the predefined thresholds θ 1 and θ 2 , respectively, then the candidateB t passes the verification.…”

Section: Feature Calculation With Feature Poolmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations