The Thermal Infrared Visual Object Tracking VOT-TIR2016 Challenge Results

Felsberg, Michael; Kristan, Matej; Matas, Jiřı́; Leonardis, Aleš; Pflugfelder, Roman; Hager, G; Berg, Amanda; Eldesokey, Abdelrahman; Ahlberg, Jörgen; Čehovin, Luka; Vojíř, Tomáš; Lukežič, Alan; Fernández, Gustavo J.; Petrosino, Alfredo; García-Martín, Álvaro; Montero, Andrés Solís; Varfolomieiev, Anton; Erdem, Aykut; Han, Bohyung; Chang, Chang-Ming; Du, Dawei; Erdem, Erkut; Khan, Fahad Shahbaz; Porikli, Fatih; Zhao, Fei; Bunyak, Filiz; Battistone, Francesco; Zhu, Gao; Seetharaman, Guna; Li, Hongdong; Qi, Honggang; Bischof, Horst; Possegger, Horst; Nam, Hyeonseob; Valmadre, Jack; Zhu, Jianke; Feng, Jiayi; Lang, Jochen; Martinez, José M.; Palaniappan, Kannappan; Lebeda, Karel; Gao, Ke; Mikolajczyk, Krystian; Wen, Longyin; Bertinetto, Luca; Poostchi, Mahdieh; Maresca, Mario Edoardo; Danelljan, Martin; Arens, Michael; Tang, Ming; Baek, Mooyeol; Fan, Nana; Al-Shakarji, Noor M.; Mikšík, Ondřej; Akin, Osman; Torr, Philip H. S.; Huang, Qingming; Martín-Nieto, Rafael; Pelapur, Rengarajan; Bowden, Richard; Laganière, Robert; Krah, Sebastian B.; Li, Shengkun; Yao, Shizeng; Hadfield, Simon; Lyu, Siwei; Becker, Stefan; Golodetz, Stuart; Hu, Tao; Mauthner, Thomas; Santopietro, Vincenzo; Li, Wenbo; Hübner, Wolfgang; Li, Xin; Li, Yang; Xu, Zhan; He, Zhenyu

doi:10.1007/978-3-319-48881-3_55

Cited by 49 publications

(39 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In both cases, we leave out around 10% of the videos during training as validation set. We evaluate our TIR tracker on the VOT-TIR2017 dataset [6], which is identical to VOT-TIR2016 dataset [5] as the 2016 edition of this benchmark was far from being saturated. It contains 25 TIR videos of varying image resolution, with an average sequence length of 740 frames adding up to a total of 13,863 frames.…”

Section: A Datasetsmentioning

confidence: 99%

“…We follow the measures and evaluation protocol proposed by the VOT-TIR2017 benchmark [5]. The two primary measures are accuracy (A) and robustness (R), which have been shown to be highly interpretable and only weakly correlated [64].…”

Section: B Evaluation Measures and Protocolmentioning

confidence: 99%

“…Recently, specialized tracking subproblems have emerged. Among these is the field of tracking in thermal infrared (TIR) images, whose importance is further increasing due to improvements of thermal infrared sensors in resolution and quality [4], [5], [6]. The advantage of thermal images is that they are not influenced by the illumination variations and shadows, and objects can be distinguished from the background as the background is normally colder.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Synthetic Data Generation for End-to-End Thermal Infrared Tracking

Zhang

González-García

Weijer

et al. 2019

IEEE Trans. on Image Process.

Self Cite

141

View full text Add to dashboard Cite

The usage of both off-the-shelf and end-to-end trained deep networks have significantly improved performance of visual tracking on RGB videos. However, the lack of large labeled datasets hampers the usage of convolutional neural networks for tracking in thermal infrared (TIR) images. Therefore, most state of the art methods on tracking for TIR data are still based on handcrafted features.To address this problem, we propose to use image-to-image translation models. These models allow us to translate the abundantly available labeled RGB data to synthetic TIR data. We explore both the usage of paired and unpaired image translation models for this purpose. These methods provide us with a large labeled dataset of synthetic TIR sequences, on which we can train end-to-end optimal features for tracking. To the best of our knowledge we are the first to train end-to-end features for TIR tracking. We perform extensive experiments on VOT-TIR2017 dataset. We show that a network trained on a large dataset of synthetic TIR data obtains better performance than one trained on the available real TIR data. Combining both data sources leads to further improvement. In addition, when we combine the network with motion features we outperform the state of the art with a relative gain of over 10%, clearly showing the efficiency of using synthetic data to train end-to-end TIR trackers.

show abstract

Section: A Datasetsmentioning

confidence: 99%

Section: B Evaluation Measures and Protocolmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Synthetic Data Generation for End-to-End Thermal Infrared Tracking

Zhang

González-García

Weijer

et al. 2019

IEEE Trans. on Image Process.

Self Cite

141

View full text Add to dashboard Cite

show abstract

“…In this subsection, we present the evaluation results on the thermal infrared object tracking benchmark VOT-TIR2016 [17] to illustrate our proposed method achieves the impressive results against most state-of-the-art trackers. We employ accuracy, robustness and EAO score as the evaluation metrics to conduct the experiments.…”

Section: Evaluation On Vot-tir2016mentioning

confidence: 99%

“…Compared trackers. For more comprehensive evaluations, we compare our LMSCO with nine state-of-the-art trackers on VOT-TIR2016 benchmark, including MDNet [16], deepMKCF [34], SHCT [15], DSST [5], NSAMF [27], S-RDCF [6], Staple [14], FCT [17] and GGT2 [35]. All these Fig.…”

Section: Evaluation On Vot-tir2016mentioning

confidence: 99%

Large Margin Structured Convolution Operator for Thermal Infrared Object Tracking

Gao

Song

et al. 2018

2018 24th International Conference on Pattern Recognition (ICPR)

View full text Add to dashboard Cite

Compared with visible object tracking, thermal infrared (TIR) object tracking can track an arbitrary target in total darkness since it cannot be influenced by illumination variations. However, there are many unwanted attributes that constrain the potentials of TIR tracking, such as the absence of visual color patterns and low resolutions. Recently, structured output support vector machine (SOSVM) and discriminative correlation filter (DCF) have been successfully applied to visible object tracking, respectively. Motivated by these, in this paper, we propose a large margin structured convolution operator (LM-SCO) to achieve efficient TIR object tracking. To improve the tracking performance, we employ the spatial regularization and implicit interpolation to obtain continuous deep feature maps, including deep appearance features and deep motion features, of the TIR targets. Finally, a collaborative optimization strategy is exploited to significantly update the operators. Our approach not only inherits the advantage of the strong discriminative capability of SOSVM but also achieves accurate and robust tracking with higher-dimensional features and more dense samples. To the best of our knowledge, we are the first to incorporate the advantages of DCF and SOSVM for TIR object tracking. Comprehensive evaluations on two thermal infrared tracking benchmarks, i.e. VOT-TIR2015 and VOT-TIR2016, clearly demonstrate that our LMSCO tracker achieves impressive results and outperforms most state-of-the-art trackers in terms of accuracy and robustness with sufficient frame rate.

show abstract