Robust Estimation of Similarity Transformation for Visual Object Tracking

Authors: Yang Li, Jianke Zhu, Steven C.H. Hoi, Wenjie Song, Zhefeng Wang, Hantang Liu8666-8673

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results demonstrate that the proposed tracker achieves very promising prediction performance compared with the state-of-the-art visual object tracking methods while still retaining the advantages of high efficiency and simplicity in conventional correlation filter-based tracking methods. In this section, we conduct four different experiments to evaluate our proposed tracker LDES comprehensively.
Researcher Affiliation Collaboration Yang Li,1 Jianke Zhu,*1,3 Steven C.H. Hoi,2 Wenjie Song,1 Zhefeng Wang,1 Hantang Liu1 1Colleage of Computer Science and Technology, Zhejiang University, Hangzhou, China 2School of Information Systems, Singapore Management University, Singapore 3Alibaba-Zhejiang University Joint Research Institute of Frontier Technologies
Pseudocode No No structured pseudocode or algorithm blocks explicitly labeled as such were found.
Open Source Code Yes The main contributions of our work are summarized as follows: ... 3) a new approach for scale and rotation estimation with efficient implementation which can improve a family of existing correlation filter-based trackers (our implementation is available at https://github.com/ihpdep/LDES).
Open Datasets Yes We first evaluate our proposed log-polar based scale estimator on OTB-2013 and OTB-100 dataset (Wu, Lim, and Yang 2013; 2015). To better evaluate our proposed approach to rotation estimation, we conduct an additional experiment on POT benchmark (Liang, Wu, and Ling 2018).
Dataset Splits No The paper uses standard benchmarks (OTB-2013, OTB-100, POT) but does not explicitly describe the training, validation, or test splits used for these benchmarks within the text.
Hardware Specification Yes All the methods were implemented in Matlab and the experiments were conducted on a PC with an Intel i7-4770 3.40GHz CPU and 16GB RAM.
Software Dependencies No The paper states 'All the methods were implemented in Matlab' but does not provide a version number for Matlab or any other specific software libraries or dependencies with their versions.
Experiment Setup Yes Experimental Settings All the methods were implemented in Matlab and the experiments were conducted on a PC with an Intel i7-4770 3.40GHz CPU and 16GB RAM. We employ Ho G feature for both translational and scale-rotation estimation, and the extra color histogram is used to estimate translational. All patch is multiplied a Hann window as suggested in (Bolme et al. 2010). η is 0.15 and λ is set to 1e 4. λφ and λα are both set to 0.01. λω is 0.015. The size of learning patch D is 2.2 larger than the original target size. Moreover, the searching window size N is about 1.5 larger than the learning patch size D. For scale-rotation estimation, the phase correlation sample size is about 1.8 larger than the original target size. All parameters are fixed in the following experiments.