Exploiting the Circulant Structure of Tracking-by-Detection with Kernels

By João F. Henriques, Rui Caseiro, Pedro Martins, Jorge P. Batista. European Conference on Computer Vision, 2012

2,356 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/5b4e50860d61095bb5fb65eaa367b131923917be

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Visual Tracking via Adaptive Spatially-Regularized Correlation Filters (2019, methodology): “The CSK [17] tracker introduces kernel trick into the CF formula.”
Beyond Correlation Filters: Learning Continuous Convolution Operators for Visual Tracking (2016, methodology): “, including scale estimation [9,31], non-linear kernels [23,24], a long-term memory [36], and by alleviating the periodic effects of circular convolution [11,15,18].”
A Novel Performance Evaluation Methodology for Single-Target Trackers (2015, methodology): “and CSK tracker [64]), NCC based tracker with motion model”
High-Speed Tracking with Kernelized Correlation Filters (2014, neutral): “A comprehensive review of tracking-by-detection is outside the scope of this article, but we refer the interested reader to two excellent and very recent surveys [1], [2].”
Correlation filters with limited boundaries (2014, methodology): “MOSSE KMOSSE MILTrack STRUCK OAB(1) SemiBoost FragTrack Our method [5] [13] [3] [11] [9] [10] [1] FaceOcc1 {1.”
Online Object Tracking: A Benchmark (2013, methodology): “Among the top 10 trackers, CSK has the highest speed where the proposed circulant structure plays a key role.”
An Exploration of Parameter Redundancy in Deep Networks with Circulant Projections (2015, background): “The properties of circulant matrices have also been exploited to avoid expensive rounds of hard negative mining in training of object detectors [14] and for real-time tracking [15].”
RGB-T Object Tracking: Benchmark and Baseline (2018, methodology): “For example, the CSK tracker [43] uses the kernel trick to improve performance.”
CREST: Convolutional Residual Learning for Visual Tracking (2017, methodology): “Typically, the convolution operation between the correlation filter W and the input X is formulated into a dot product in the Fourier domain [17, 31, 18].”
LaSOT: A High-Quality Benchmark for Large-Scale Single Object Tracking (2018, neutral): “We evaluate 35 algorithms on LaSOT to provide extensive baselines, comprising deep trackers ( e.g ., MDNet [42], TRACA [5], CFNet [50], SiamFC [4], StructSiam [59], DSiam [16], SINT [49] and VITAL [48]), correlation ﬁl-ter trackers with hand-crafted features ( e.g ., ECO HC [7], DSST [8], CN [11], C”
GOT-10k: A Large High-Diversity Benchmark for Generic Object Tracking in the Wild (2018, methodology): “CSK introduces non-linear kernels to correlation ﬁlters.”
Multi-task Correlation Particle Filter for Robust Object Tracking (2017, methodology): “Correlation filters have recently been introduced into visual tracking and shown to achieve high speed as well as robust performance [4, 9, 16, 15, 18, 26, 24, 21, 25, 29].”
The Visual Object Tracking VOT2016 Challenge Results (2016, methodology): “Without loss of generality, conventional CN tracker [87] is utilized as a sample CFT baseline to expeditiously track each target object part by feeding color-induced attributes into fast CSK tracker [99].”
Triplet Loss in Siamese Network for Object Tracking (2018, background): “For example, Henriques et al. [12] applied the Histogram of Oriented Gradients (HOG) feature instead of gray feature in [11] to achieve more robust tracking performance.”
3D Pose from Motion for Cross-View Action Recognition via Non-linear Circulant Temporal Encoding (2014, neutral): “Furthermore, temporal consistency is enforced by interpreting filtered circular convolution as a Regularized Least Squares problem, and solving it efficiently in the frequency domain [7].”