3D face (facial expressions) databases

Human detection

  • Dataset and ground-truth
    • [PETS 2009 TUD EPFL](http://www.milanton.de/data.html)
  • Papers:
    • Piotr Dollár’s work
      • 09 BMVC - Integral Channel Features
      • 10 BMVC - The Fastest Pedestrian Detector in the West
      • 12 ECCV - Crosstalk Cascades for Frame-Rate Pedestrian Detection
      • 12 PAMI - Pedestrian Detection: An Evaluation of the State of the Art
      • 14 PAMI - Fast Feature Pyramids for Object Detection
  • Other:

Human Pose Estimation

Object tracking

Saliency detection

Image feature detection and description 1

Keypoint detectors

  • FAST: high-speed corner detector, in OpenCV
  • AGAST: even faster than the FAST. A multi-scale version of this method is used for the BRISK descriptor (10 ECCV).

Binary descriptors

  • BRIEF – fast and accurate interest point descriptor (not invariant to rotations and scale)
  • ORB – Oriented-Brief descriptor (invariant to rotations, but not scale), in OpenCV
  • BRISK – efficient binary descriptor invariant to rotations and scale, in OpenCV
  • FREAK – faster than BRISK (invariant to rotations and scale), in OpenCV

SIFT & SURF Implementations

Other Local Feature Detectors and Descriptors

Global Image Descriptors

  • GIST: MATLAB C++ (Lear’s version)
  • CENTRIST (in libHIK) – global visual descriptor for scene categorization and object detection. Another version can be found here.

Image deblur

Image Dehaze

Steoro matching


  1. Source page of SYSU: http://gitl.sysu.edu.cn/downloads.