190330Mapillary_Meetup_Nara

* 瀬戸寿一 (2019年3月): Mapillaryを支えるデータ＆アルゴリズム＆チャレンジ、Mapillary Meetup 2019 Nara, Kansai、奈良女子大学、奈良県奈良市、2019年3月30日。

Transcript

1. / 3 /9 11 ,/9/ / /3
   @
   O MS bS a K
   )1 N )1 (2/9 19 1 019
   1 911 / K KC GK
   

   View Slide

2. ,, /3 12 / 0
   - 01 2C
   S Y 2 4
   4 8810
   a 1 4
   690 kmM C2A
   B:
   

   View Slide

3. ,, /3 12 / 0
   https://mapillary.github.io/mapillary_greenhouse/global-challenge/q2-2018/
   Score = (total unique KM contributed + (images/1000)) * (1 + (count of local participants/10))
   

   View Slide

4. /, 04 23/ 0 1 ,
   #
   https://mapillary.github.io/mapillary_greenhouse/
   

   View Slide

5. /, 0 23/ 0 1 ,
   4 4
   

   View Slide

6. /, 0 23/ 0 1 ,
   8
   https://www.johnscreekga.gov/NewsAndEvents/News-Archive/2017-
   News/Complete-the-Map-residents-can-participate-in-str
   

   View Slide

7. , /11, 9 7 , , ,2 ,/
   
   
   • 0 er , /11, 9p
   • ,2/ ,7/32 nMo e k e M
   , /11, 9 u
   M TM i
   • O N t Kg y lz
   sa
   https://blog.mapillary.com/update/2018/01/26/youth-citizen-engagement-school-routes-reporting.html
   

   View Slide

8. /, 0 23/ 0 1 ,
   Mapillary images are available under the Creative Commons Attribution-ShareAlike 4.0 (CC BY-SA 4.0)
   

   View Slide

9. / 455/9 11 ,/9/ / :/4
   9 A
   sP ijO Srq
   • NP K
   • 7 . 3 7 C / A
   • A A
   – . 3 L .e G m _fI I b
   g MU C c acd 9C h ln
   CC 79C 3 9 A C3 9 7
   • M L
   • C A 7 C 3C9
   • : 3C
   – 9/2240 :43 : T
   – 4 : T
   – 4 1: T
   • . 7 C C9 op
   

   View Slide

10. ,, /3 12 / 0
    
    
    http://openaccess.thecvf.com/content_ICCV_2017/papers/Neuhold_The_Mapillary_Vi
    stas_ICCV_2017_paper.pdf
    

    View Slide

11. ,, /3 12 / 0
    1 5 2 9
    8 9 4
    9 201
    The Mapillary Vistas Dataset for Semantic Understanding of Street Scenes
    doi: 10.1109/ICCV.2017.534
    

    View Slide

12. ,, /3 12 / 0
    

    View Slide

13. ,, /3 12 / 0
    

    View Slide

14. /, 04 23/ 0 1 ,
    - - - - ( -
    ( - - )
    https://research.mapillary.com/lsun.html
    

    View Slide

15. /, 0 23/ 0 1 ,
    0 2 1 ,
    •Road markings: increased from 6 classes to 21 classes including speed limits and direction
    arrows, which are essential classes for building navigation and HD maps.
    •Driveway: a new class that we’ve introduced to help enhance data extraction for last-mile
    •Barriers: increased from 6 to 11 classes. Detailed refinement of the barrier classes is the first
    step to enhance the perception capacity for autonomous driving under different scenarios.
    •Signages: the billboard class is now refined to advertisement, storefront, and general
    information signages. This facilitates automatic and in-depth location-aware data extraction
    for marketing, business, and point-of-interest analysis.
    •Traffic light states: dynamic properties are associated with static objects. In this case, we
    have refined 100K traffic light annotations with different state
    

    View Slide

16. /, 0 23/ 0 1 ,
    http://www.robustvision.net/
    

    View Slide

17. /, 0 23/ 0 1 ,
    https://blog.mapillary.com/update/2018/06/14/robust-cv.html
    

    View Slide

18. /, 0 23/ 0 1 ,
    http://www.wilddash.cc/benchmark/resultsvis_rob18_sem?hc=semantic_rob
    

    View Slide

19. ,, /3 12 / 0
    ( 2 0 2 9
    • 3 CDo C yz p s
    j g
    • u e Waf b
    1 2 2 c lnPviM
    DP O P z M t
    r M DP
    https://blog.mapillary.com/update/2019/02/13/announcing-capture-projects.html
    

    View Slide

20. ,, /3 12 / 0
    Mapillary Vistas Dataset
    
    25,000 & 152
    https://research.mapillary.com/
    

    View Slide

21. ,, /3 12 / 0
    PD Tn
    • clu a Pn T R ie
    • y r PD T p o Ny
    m I A NI
    s
    • N C t Iy SV hI c
    l Gy g T AM
    n
    https://blog.mapillary.com/update/2019/03/12/toyota-research-institute-and-mapillary.html
    

    View Slide

22. ,, /3 12 / 0
    3 - 2 09 9CC 0 9 9 09: 9
    • P tp e c a S [ uy nrNi
    s SWV b S R o[ S N g
    l m % S N SWV
    • % 1./ ]dh
    • C : E 9 % % E :
    C9 9CC C 9 9 C9: 9
    •
    9 N K
    M
    

    View Slide

23. ,, /3 12 / 0
    ! : 2 7 : ey Eic n
    • b MP [r n W
    • nj M EN ] [ac g
    • : / O V
    :3 7 7 ft eoMp lso
    https://blog.mapillary.com/update/2019/03/27/announcing-verification-projects.html
    

    View Slide

24. /, 04 23/ 0 1 ,
    Thank you!
    tosseto@csis.u-tokyo.ac.jp
    http://researchmap.jp/tosseto
    https://speakerdeck.com/tosseto
    

    View Slide