Upgrade to Pro — share decks privately, control downloads, hide ads and more …

三次元形状とディープラーニング

 三次元形状とディープラーニング

三次元形状処理とディープラーニングの初歩についてまとめたスライドです。

Tatsuya Yatagawa

December 21, 2020
Tweet

More Decks by Tatsuya Yatagawa

Other Decks in Technology

Transcript


  1. 機械学習
    ニューラルネット
    深層学習

    View full-size slide



  2. • θ

    = ; = 2
    2 + 1
    + 0

    = arg min ෍
    =0
    −1
    (
    ,
    ; )

    View full-size slide





  3. 0
    = 0
    + 0
    1
    = 1
    + 1
    2
    = 2
    + 2
    = 0
    0
    + 1
    1
    + 2
    2
    +
    = 0
    0
    + 1
    1
    + 2
    2

    +(0
    0
    + 1
    1
    + 2
    2
    + )

    View full-size slide





  4. 0

    1

    2

    0
    = ReLU(0
    + 0
    )
    1
    = ReLU(1
    + 1
    )
    2
    = ReLU(2
    + 2
    )



    View full-size slide

  5. = 32 = 128 = 512


    View full-size slide



  6. = 512
    = 1 = = 4

    View full-size slide












  7. View full-size slide

  8. (c) Velodyne
    (c) Andrew Tallon
    (c) Microsoft
    (c) Sony

    View full-size slide

  9. ′ =


    ′ = +




    View full-size slide










  10. , , = 0

    View full-size slide









  11. , , = 0

    View full-size slide






  12. View full-size slide





  13. View full-size slide





  14. View full-size slide








  15. View full-size slide









  16. , , = 0

    View full-size slide




  17. (c) Velodyne

    View full-size slide




  18. (0
    , 0
    , 0
    )
    (1
    , 1
    , 1
    )
    (
    ,
    ,
    )
    (−1
    , −1
    , −1
    )
    (+1
    , +1
    , +1
    )
    (0
    , 0
    , 0
    )
    (1
    , 1
    , 1
    )
    (−1
    , −1
    , −1
    )
    (+1
    , +1
    , +1
    )
    (
    ,
    ,
    )

    View full-size slide









  19. View full-size slide









  20. , , = 0

    View full-size slide





  21. View full-size slide



  22. ∗ = ℱ−1(ℱ ⋅ ℱ )
    ′ = ℱ−1(Θ ∘ ℱ )

    View full-size slide


  23. =
    0 1 1 0
    1 0 1 0
    1 1 0 1
    0 0 1 0
    = − −
    1
    2−
    1
    2

    = ෍



    :

    View full-size slide




  24. = T

    View full-size slide

  25. • ()
    ′ = ℱ−1(Θ ∘ ℱ )
    ′ =

    ℱ = ()
    ℱ−1 = ()

    View full-size slide








  26. +1

    +2


    +1 = ReLU ⋅ ෍



    View full-size slide







  27. View full-size slide









  28. , , = 0

    View full-size slide





  29. View full-size slide





  30. View full-size slide





  31. View full-size slide





  32. View full-size slide





  33. View full-size slide









  34. View full-size slide










  35. View full-size slide