Correlation Filter Trackers notes

The notes are ordered by year ascendant. Tracker listed in each year are in no particular order

Conferences names:

• CVPR = IEEE Conference on Computer Vision and Pattern Recognition
• ECCV = European Conference on Computer Vision
• ICCV = International Conference on Computer Vision
• IJCV = International Journal of Computer Vision
• TPAMI = IEEE Transactions on Pattern Analysis and Machine Intelligence

2018

• MKCFup

  • Paper: "High-speed Tracking with Multi-kernel Correlation Filters"
  • Presented in: arXiv
  • Major Contribution/s: Introduce the multi-kernel learning (MKL) into KCF in a different way compared to the MKCF tracker, achieving better performance with not much drop in fps.
  • Code: -

• LCT Tracker v2

  • Paper: "Adaptive Correlation Filters with Long-Term and Short-Term Memory for Object Tracking"
  • Presented in: ICJV
  • Major Contribution/s: Learn multiple adaptive correlation filters with both long-term memory and short-term memory of target appearance for robust object tracking.
  • Code: Matlab

2017

• ACFN

  • Paper: "Attentional Correlation Filter Network for Adaptive Visual Tracking"
  • Presented in: CVPR
  • Major Contribution/s: Propose a new framework in which an attention network that selects the best module (a set of Correlation Filters) to track the object in a certain frame.
  • Code: Matlab

• CFNet (Project Page)

  • Paper: "End-to-end representation learning for Correlation Filter based tracking"
  • Presented in: CVPR
  • Major Contribution/s: First work to propose training the Correlation Filter jointly with a Siamese Network in an end-to-end fashion.
  • Code: Matlab

• DCFNet

  • Paper: "DCFNet: Discriminant Correlation Filters Network for Visual Tracking"
  • Presented in: arXiv
  • Major Contribution/s: Another work to propose training the Correlation Filter jointly with a Siamese Network. Different from CFNet in network architecture and place of the Correlation Filter in the Tracker architecture.
  • Code: Matlab | Python (PyTorch)

2015

• MKCF

  • Paper: "Multi-kernel Correlation Filter for Visual Tracking"
  • Presented in: ICCV
  • Major Contribution/s: Introduce multi-kernel to correlation filter based trackers.
  • Code: .exe file

• KCF (Project Page)

  • Paper: "High-Speed Tracking with Kernelized Correlation Filters"
  • Presented in: TPAMI
  • Major Contribution/s: Derive a new Kernelized Correlation Filter (KCF) and propose a fast multi-channel extension of linear correlation filters, via a linear kernel, called by authors Dual Correlation Filter (DCF).
  • Code: Matlab | More codes in Project page

• LCT Tracker v1

  • Paper: "Long-Term Correlation Tracking"
  • Presented in: CVPR
  • Major Contribution/s: Introduced online random fern classifier as re-detection component for long-term tracking.
  • Code: Matlab

• CF2 (Project Page)

  • Paper: "Hierarchical Convolutional Features for Visual Tracking"
  • Presented in: ICCV
  • Major Contribution/s: Learn a Correlation Filter on each Convolutional Layer of a pre-trained network, taking advantage of the different information of features from different layers.
  • Code: Matlab

2012

• CSK (Project Page)
  • Paper: "Exploiting the Circulant Structure of Tracking-by-detection with Kernels"
  • Presented in: ECCV
  • Major Contribution/s: Introduced Ridge Regression problem with circulant matrix to apply kernel methods.
  • Code: Matlab | More codes in Project page

2010

• MOSSE (Minimum Output Sum of Squared Errors)
  • Paper: "Visual Object Tracking using Adaptive Correlation Filters"
  • Presented in: CVPR
  • Major Contribution/s: Pioneering work in introducing Correlation Filters for visual tracking. Filter is single channel.
  • Code: -