Riemannian manifolds, kernels and learning

Поділитися
Вставка
  • Опубліковано 25 лип 2016
  • I will talk about recent results from a number of people in the group on Riemannian manifolds in computer vision. In many Vision problems Riemannian manifolds come up as a natural model. Data related to a problem can be naturally represented as a point on a Riemannian manifold. This talk will give an intuitive introduction to Riemannian manifolds, and show how they can be applied in many situations. Examples that will be considered are the Essential manifold, relevant in structure from motion; the manifold of Positive Definite matrices and the Grassman Manifolds, which have a role in object recognition and classification, and the Kendall shape manifold, which represents the shape of 2D objects
  • Наука та технологія

КОМЕНТАРІ • 48

  • @JyujinPlus
    @JyujinPlus 4 роки тому +20

    “Start slow so you’re not lost from slide one.”
    You, sir, are my hero

  • @bartholomeosphinx4382
    @bartholomeosphinx4382 6 років тому +85

    Same problem as with all Microsoft Research presentations - the producer of the film is ignorant as to the importance of the slides.

  • @davidk9382
    @davidk9382 4 роки тому +6

    Thank you to you and your students for sharing this.

  • @brendawilliams8062
    @brendawilliams8062 2 роки тому

    Thankyou. The Professor insight into triangulation was appreciated.

  • @wananajakbandit
    @wananajakbandit 7 років тому +55

    The content is great, but the production of this video is infuriating. Please leave the slide up for long enough for us to read the slide. As it is, you show the slide for a second, and then switch to a different camera angle.

    • @vector8310
      @vector8310 6 років тому +17

      That's why God created the pause button

    • @davidk9382
      @davidk9382 4 роки тому +2

      @@vector8310 was about to say "boomer", but that would have been harsh.

    • @miguelcerna7406
      @miguelcerna7406 3 роки тому +2

      . who likes doing that? Ruins the flow of the lecture. Everyone has tje exact issue

    • @miguelcerna7406
      @miguelcerna7406 3 роки тому

      To everyone complaining about the slides...it gets worse, sound goes off at around min 29.

    • @pinklady7184
      @pinklady7184 3 роки тому

      I understand your frustration. Pause button helps.

  • @neoneo1503
    @neoneo1503 Рік тому +4

    14:42 The back and forth between Tangent space and manifold (iteration algorithm on manifold - Weiszfeld algorithm)

  • @patrickjames4245
    @patrickjames4245 2 роки тому

    I just set someone up for the 1,000th like. Congrats to the Richard Hartley and the Microsoft Research team for creating this video. Very successful.

  • @daleowens7695
    @daleowens7695 4 роки тому +10

    A bit beyond me, but this must be the theoretical underpinnings of how they produced the 3d graphics of landscapes from satellite images for MS Flight Simulator 2020.

  • @gokulrp9064
    @gokulrp9064 3 роки тому

    how come the geodesic distance of the first example(the sphere shown in the corner ) comes like that

  • @tanmayjaiswal5935
    @tanmayjaiswal5935 2 роки тому +4

    Him: "The tangent plane is in fact the tangent plane".
    Me: Hmmm, yes. It do be that way...

  • @Diego-es9yb
    @Diego-es9yb 3 роки тому +5

    im here listening but i dont understand anything

  • @ILikeWeatherGuy
    @ILikeWeatherGuy 7 років тому +14

    called exponential map because the trancendential e^x is the same when integrating/differentiating.

    • @tobiaszb
      @tobiaszb 2 роки тому

      Also in Lie Groups, the series definition of the exponent map holds ^^.
      "in the setting of matrix Lie groups, the exponential map is the restriction of the matrix exponential to the Lie algebra." en.wikipedia.org/wiki/Exponential_map_(Lie_theory)

  • @vegetableball
    @vegetableball 5 років тому +11

    Suggestion: Speaker's name should be in the description.

    • @remidelmas9286
      @remidelmas9286 5 років тому +1

      the speaker is richard hartley www.microsoft.com/en-us/research/video/riemannian-manifolds-kernels-and-learning/

    • @afaisaladhamshaazi7519
      @afaisaladhamshaazi7519 4 роки тому +1

      I thought it was Kurtwood Smith

  • @therealkalashnikov5460
    @therealkalashnikov5460 6 років тому +10

    Math is the best :-) though I never received a passing grade.

  • @amirdaneshmand9743
    @amirdaneshmand9743 6 років тому +13

    Please show the slides not the lecturer

  • @ProfessionalTycoons
    @ProfessionalTycoons 5 років тому +3

    such mathematical beauty

  • @Alley00Cat
    @Alley00Cat 6 років тому +5

    The necessary slides are here: www.robots.ox.ac.uk/~vgg/rg/slides/Oxford-Mar-2014.pdf

  • @Acheiropoietos
    @Acheiropoietos 9 днів тому

    Is no one going to mention his choice of shirt and cardigan? They certainly don’t match as well as the manifold projections.

  • @mahdihamad9545
    @mahdihamad9545 5 років тому +1

    links to slides www.robots.ox.ac.uk/~vgg/rg/slides/Oxford-Mar-2014.pdf

    • @bonbonpony
      @bonbonpony Рік тому +2

      If you also add the login and password, you'll get another like.

  • @davidk9382
    @davidk9382 4 роки тому +1

    38K views? That cant be real.

    • @davidk9382
      @davidk9382 4 роки тому

      The hand that draws itself.

  • @davidwilkie9551
    @davidwilkie9551 5 років тому +4

    If I "had the language", the direct relationships between e-Pi-i temporal resonance points on a zero axis harmonic, normal to the exponential map.., tangential vector spacing, would be "obvious", but these basic elements of the Quantum Operator, aligned on coaxial cones or sheaves, are understood empirically in terms of spacetime first, and to an observation of cause-effect, the reverse process of the Quantum Fields Modulation Mechanism is harmonically transparent. The observable Origin of QM-Time modulation in-form-ation is the Universal context of macro-micro vanishing point singularity connection.., "insideout", inflating the image-universe spectrum of time duration temporal superposition, eternity-now.
    It's not a Big Bang, but it looks like it superficially, in elemental statements, the "i-reflection" history or "Echo-location" positioning of QM-TIMESPACE.., Mathematically.
    The Observable Universe is WYSIWYG.., When inside the loops of time duration, at the Node of QM-Time eternity-now singularity connection. The time duration loops surrounding the combined vanishing point node of Observation/Origin are the sum-of-all-history here-now image, and that's the ordinary existence we've always known intuitively, but has been lost in the obscurity of a superficial narrative overlay.

    • @tanmayjaiswal5935
      @tanmayjaiswal5935 2 роки тому +4

      Looks like someone trained GPT-1 on Math and physics textbooks...
      You completely lost me bud, but I don't think the comment was intended for someone like me to begin with 😂
      Carry on...

  • @forheuristiclifeksh7836
    @forheuristiclifeksh7836 Місяць тому +1

    7:00

  • @forheuristiclifeksh7836
    @forheuristiclifeksh7836 Місяць тому +1

    4:15

  • @forheuristiclifeksh7836
    @forheuristiclifeksh7836 Місяць тому

    44:39

  • @thomasolson7447
    @thomasolson7447 7 місяців тому

    Seems like math that got out of hand and is useful to no one. I think I'll stick to the SoME videos.

  • @juliensorel1427
    @juliensorel1427 4 роки тому +4

    Really bad .. going from manifolds ...with basically no examples to Hilbert Space inner product .... ?????