Online Object Tracking: A Benchmark
read more
Citations
High-Speed Tracking with Kernelized Correlation Filters
Object Tracking Benchmark
Fully-Convolutional Siamese Networks for Object Tracking
Accurate scale estimation for robust visual tracking
ECO: Efficient Convolution Operators for Tracking
References
Histograms of oriented gradients for human detection
Robust Real-Time Face Detection
An iterative image registration technique with an application to stereo vision
Robust real-time face detection
C ONDENSATION —Conditional Density Propagation forVisual Tracking
Related Papers (5)
Frequently Asked Questions (18)
Q2. What is the way to evaluate the performance of the tracking algorithms?
The authors use the precision plots based on location error metric and the success plots based on the overlap metric, to analyze the performance of each algorithm.
Q3. What is the recent development in object tracking?
the discriminative model has been widely adopted in tracking [15, 4], where a binary classifier is learned online to discriminate the target from the background.
Q4. What are some of the popular features in tracking algorithms?
In addition to template, many other visual features have been adopted in tracking algorithms, such as color histograms [16], histograms of oriented gradients (HOG) [17, 52], covariance region descriptor [53, 46, 56] and Haar-like features [54, 22].
Q5. What methods have been adapted to the tracking problem?
Numerous learning methods have been adapted to the tracking problem, such as SVM [3], structured output SVM [26], ranking SVM [7], boosting [4, 22], semiboosting [23] and multi-instance boosting [5].
Q6. What are the two tests that are commonly used in the real world?
The proposed test scenarios happen a lot in the realworld applications as a tracker is often initialized by an object detector, which is likely to introduce initialization errors in terms of position and scale.
Q7. What is the purpose of this work?
In this work, the authors build a code library that includes most publicly available trackers and a test dataset with ground-truth annotations to facilitate the evaluation task.
Q8. What is the conventional way to evaluate a tracker?
The conventional way to evaluate trackers is to run them throughout a test sequence with initialization from the ground truth position in the first frame and report the average precision or success rate.
Q9. What is the performance of the affine motion trackers?
The results show that trackers with affine motion models (e.g., ASLA and SCM) often handle scale variation better than others that are designed to account for only translational motion with a few exceptions such as Struck.
Q10. What is the overlap score for the tracked bounding box rt?
Given the tracked bounding box rt and theground truth bounding box ra, the overlap score is defined as S = | rt⋂ ra || rt ⋃ ra |, where ⋂ and ⋃ represent the intersec-tion and union of two regions, respectively, and | · | denotes the number of pixels in the region.
Q11. What are the common sources of tracking source codes?
more tracking source codes have been made publicly available, e.g., the OAB [22], IVT [47], MIL [5], L1 [40], and TLD [31] algorithms, which have been commonly used for evaluation.
Q12. What is the main reason for the Haar-like features?
This indicates that the Haar-like features are somewhat robust to background clutters due to the summation operations when computing features.
Q13. How can the authors use the background information in the discriminative model?
It can be exploited by using advanced learning techniques to encode the background information in the discriminative model implicitly (e.g., Struck), or serving as the tracking context explicitly (e.g., CXT).
Q14. What is the main issue in evaluating tracking algorithms?
One common issue in assessing tracking algorithms is that the results are reported based on just a few sequences with different initial conditions or parameters.
Q15. How do the authors measure the performance of a tracker?
the authors propose two ways to analyze a tracker’s robustness to initialization, by perturbing the initialization temporally (i.e., start at different frames) and spatially (i.e., start by different bounding boxes).
Q16. Why are the plots of SRE presented for analysis?
Due to the space limitation, the plots of SRE are presented for analysis in the following sections, and more results are included in the supplement.
Q17. What is the performance bottleneck of the trackers?
Compared with other higher ranked trackers, the performance bottleneck of them can be attributed to their adopted representation based on sparseprincipal component analysis, where the holistic templates are used.
Q18. What is the main problem of the Lucas-Kanade algorithm?
Matthews et al. [39] address the template update problem for the Lucas-Kanade algorithm [37] where the template is updated with the combination of the fixed reference template extracted from the first frame and the result from the most recent frame.