3 minute Thingy

Transcript

1. 1
   Query
   images
   Result
   images
   Realtime learning
   Dataset: wikiart images - https://www.wikiart.org/.
   No image pre-processing performed. No textual data used.
   Feature vectors generated automatically with deep learning.
   www.xyggy.com © 2019
   

   View Slide

2. 2
   Add 2nd image to query
   www.xyggy.com © 2019
   

   View Slide

3. 3
   www.xyggy.com © 2019
   Add 3rd image to query
   

   View Slide

4. 4
   Add 4th image to query
   www.xyggy.com © 2019
   

   View Slide

5. 5
   www.xyggy.com © 2019
   Thingy
   

   View Slide

6. 6
   Add and remove images from query
   Query images
   Result images
   Dataset: wikiart images - https://www.wikiart.org/.
   No image pre-processing performed. No textual data used.
   Feature vectors generated automatically with deep learning.
   www.xyggy.com © 2019
   

   View Slide

7. 7
   Add 2nd image to query
   www.xyggy.com © 2019
   

   View Slide

8. 8
   Add two more images to query
   www.xyggy.com © 2019
   

   View Slide

9. 9
   Remove two images from query
   www.xyggy.com © 2019
   

   View Slide

10. 10
    Thingy
    www.xyggy.com © 2019
    

    View Slide

11. 11
    More like this
    Query images
    Result images
    Dataset: 110K Visual Genome images - http://visualgenome.org/
    No image pre-processing performed. No textual data used.
    Feature vectors generated automatically with deep learning.
    www.xyggy.com © 2019
    

    View Slide

12. 12
    More like this
    www.xyggy.com © 2019
    

    View Slide

13. 13
    More like these
    Query images
    Result images
    www.xyggy.com © 2019
    

    View Slide

14. 14
    More like these
    www.xyggy.com © 2019
    

    View Slide

15. 15
    Thingy
    www.xyggy.com © 2019
    

    View Slide

16. 16
    More like these, less like others
    Query images
    Result images
    Dataset: wikiart images - https://www.wikiart.org/.
    No image pre-processing performed. No textual data used.
    Feature vectors generated automatically with deep learning.
    www.xyggy.com © 2019
    

    View Slide

17. 17
    More like positives, less like negatives
    www.xyggy.com © 2019
    

    View Slide

18. 18
    More like positives, less like negatives
    www.xyggy.com © 2019
    

    View Slide

19. 19
    More like negatives, less like positives
    www.xyggy.com © 2019
    

    View Slide

20. 20
    More like negatives, less like positives
    www.xyggy.com © 2019
    

    View Slide

21. 21
    yin and yang
    www.xyggy.com © 2019
    

    View Slide

22. 22
    Thingy
    www.xyggy.com © 2019
    

    View Slide

23. 23
    Taking photos is free
    Farmer wants to know disease type affecting tomato crop.
    Photos are from the same plant.
    Dataset: 80K plant leaf images of different crops https://plantvillage.org/
    No image pre-processing performed. No textual data used.
    Feature vectors generated automatically with deep learning.
    #1 #2 #3
    www.xyggy.com © 2019
    

    View Slide

24. 24
    Query with photo #1 only
    Top 5 results:
    1 potato_early_blight
    2 tomato_early_blight
    3 potato_early_blight
    4 potato_early_blight
    5 tomato_early_blight
    2nd result identifies disease
    www.xyggy.com © 2019
    

    View Slide

25. 25
    Query with photo #2
    Top 5 results:
    1 tomato_early_blight
    2 celery_early_blight
    3 tomato_early_blight
    4 potato_early_blight
    5 tomato_early_blight
    1st result identifies disease
    www.xyggy.com © 2019
    

    View Slide

26. 26
    Query with photo #3
    Top 5 results:
    1 soybean_leaf_blight
    2 celery_early_blight
    3 potato_early_blight
    4 tomato_early_blight
    5 soybean_leaf_blight
    4th result identifies disease
    www.xyggy.com © 2019
    

    View Slide

27. 27
    Query with photo #1 and #2
    Top 5 results:
    1 tomato_early_blight
    2 tomato_early_blight
    3 celery_early_blight
    4 potato_early_blight
    5 potato_early_blight
    top 2 positions identify disease
    www.xyggy.com © 2019
    

    View Slide

28. 28
    Add photo #3 to query
    top 3 positions identify disease
    Top 5 results:
    1 tomato_early_blight
    2 tomato_early_blight
    3 tomato_early_blight
    4 potato_early_blight
    5 potato_early_blight
    www.xyggy.com © 2019
    

    View Slide

29. 29
    Thingy
    www.xyggy.com © 2019
    

    View Slide