Academic
Publications
Fast keypoint recognition using random ferns

Fast keypoint recognition using random ferns,IEEE Transactions on Pattern Analysis and Machine Intelligence,Mustafa Özuysal,Michael Calonder,Vincent L

Cumulative Annual
    • ...Here we take the approach of simulating novel viewpoints [6, 8, 15, 18] and adding features extracted from affine transformed training images to our model...
    • ...A more tractable approach to account for viewpoint change is to simulate novel viewpoints by applying affine or perspective transformations to the model images [6, 8, 18]...
    • ...Viewpoint simulation has been used to determine a keypoint’s repeatability [6] and to model a keypoint’s local appearance [8, 18]...

    Edward Hsiaoet al. Making specific features less discriminative to improve point-based 3D...

    • ...Recently, randomized learning techniques such as Random Forests [1], [2], [3] and Ferns [4] have been introduced to the vision community...

    Martin Godecet al. On-Line Random Naive Bayes for Tracking

    • ...We use FERNS as an object class detector [11], [12]...

    Alper Aydemiret al. Simultaneous object class and pose estimation for mobile robotic appli...

    • ...The AVS component uses the object recognition method proposed in [20] and the models associated with object classes reside in the categorical layer...

    Jeremy L. Wyattet al. Self-Understanding and Self-Extension: A Systems and Representational ...

    • ...A recent approach [6] introduced signatures, a local descriptor that can be computed much faster than SIFT [14] or SURF [5]: Given a Fern classifier [19] trained offline to recognize a number of keypoints extracted from an image database, the signature of a new keypoint is taken to be the response of the Fern and can be used to match it by performing a nearest neighbor search among signatures...
    • ...The signature vectors introduced in [6] describe the appearance of an image patch p in terms of the responses of a Fern classifier [19] trained offline to recognize a predefined set B of N reference keypoints...
    • ...1Note that this response is computed as the average response as in [1] and not as a Naive Bayesian score as in [19]...

    Michael Calonderet al. Compact signatures for high-speed interest point description and match...

Sort by: