Home
build details

Show: section status errors & todos local changes recent changes last change in-page changes feedback controls

Learning Approaches to VO

Modified 2019-01-07 by pravishsainath

  • Learning can be used in any stage of the odometry pipeline such as:
    • features detection
    • estimating feature correspondences
    • homography estimation


  • In general, for the motion estimation step, the problem is framed to use the labelled data of image sequences for training a regression or classification model.



  • It does not require the camera calibration parameters to be known explicitly.

  • It can estimate translation to the correct scale and are robust against similar kind of noises with which it is trained.


Several machine learning approaches have been tried and deep learning approaches are currently picking up pace. standard methods might evolve over time.



Deep VO Neural Network Model
Because of mathjax bug

Modified 2018-12-24 by Pravish Sainath

Gregory Dudek and Michael Jenkin. Computational principles of mobile robotics. Cambridge university press, 2010.    Rached Dhaouadi and A. Abu Hatab. Dynamic modeling of differential-drive mobile robots using lagrange and newton-euler methodologies: A unified framework. Advances in Robotics & Automation, 2(2):1–7, 2013.   

No questions found. You can ask a question on the website.