1. Image searching Global Brand Database
Cyberworld
20 June
2018
Sandrine Ammann
Marketing & Communications Officer

2. Agenda
What’s new?
Image search
As a user
History
Future
Quiz
Q&A

3. What’s new?
Italian marks added
Refresh of US and UEA data

4. Image search

5. Image search 2/3 steps 1. Use an image as unique search term
2. Rank search results by visual similarity to an image

6. Image = search term Pick a strategy
Message take longer: anking the entire collection of images in the Global Brand Datbase by their similarity to your source image, the sorting of the millions of images in our collection may take a few seconds to process.

7. Filter results by image: drag

8. Filter results by image: browse
Click the .browse. button and select an image file from your local computer. Please note that we can only analyze JPG, GIF, BMP and PNG files

9. Step 2: strategies

10. Strategy: shape
Shape . This comparison method looks only at the distribution of lines in your source image, and looks for similar images in our collection. Color is completely ignored for this type of search.

11. Strategy: color
Color . This comparison method looks only at the colors in the image, and how they are distributed in the image. It.s not as effective at matching shapes as the Shape or Texture methods, but is the most accurate for matching colors.

12. Strategy: texture
Texture . Similar to the .Shape. method, this comparison option ignores color completely, and also ignores the distribution of lines in the image, only paying attention to the types of lines in the image. This may be effective for some images that are not well matched by the .Shape. option

13. Strategy: composite
Composite . This option uses both color and line information in it.s analysis, but since it uses both simultaneously, it.s not as accurate as other options, but you may still receive interesting results.

14. Step 3 (optional): image type

15. Image type: verbal
Verbal . These are marks that are determined to have only a textual element of some form, without being noted as having a design element. You can find any of these marks by searching in the .Text. in the .Brand. tab in the Global Brand Database page. Note that these are only records which contain an image. If the brand is purely textual and doesn.t have an associated image, this filter will not match. For some countries, it is impossible to determine the type of image (see the .Unknown. option), so this filter will automatically exclude those images as they may or may not contain a device.

16. Image type: non verbal
Nonverbal . This will match images that are determined to have no textual element at all, but only some form of device, shape, color, scent or sound.

17. Image type: combined
Combined . This will match images associated with records that are determined to have both a textual and non-textual element.

18. Image type: unknown
Unfortunately, in some collections, it is impossible to determine what is comtained in the associated image, and these images are marked .Unknown.. They may be verbal, non-verbal or combined.

22. Image similarity: history
Released in 2014
Based on Image features:
Shape
Color
Texture
Very effective: simple geometric shapes

23. Image similarity: history

24. Image similarity: limitation
Semantic similarity

25. Limitations: example

26. Limitations: example

27. Limitations: example

28. GBD Image Search Next Steps

29. Image similarity: future
AI based search
Semantic search by definition
Automated classification of images
Release date: end of 2018

30. Image classification: example
Other cultivated areas (15.1)
Sun, rising or setting (partially exposed or partially o... (13.8)
Open sea, stretches of water without shore, multiple wav... (12.9)
Semi-circles (11.8)
Mountains, mountain landscapes (10.9)
Farms (10.2)
Rainbows (10.1)
Other representations of the sun (9.9)
Roads without lines or dividers (9.5)
Single wave of water (9.4)

31. Image classification: example
Dogs (14.8)
Stylized cats, dogs, wolves, foxes, bears, lions or tige... (14.5)
Yokes, collars, leashes, and harnesses for other animals (13.8)
Shadows or silhouettes of dogs (13.5)
Domestic cats (10.9)
Stylized elephants, hippopotami, rhinoceri, giraffes, al... (9.8)
Elephants, mammoths (9.7)
Costumed animals and those with human attributes in this... (9.6)
Hearts (9.2)
Cats, dogs, wolves, foxes, bears, lions or tigers with f... (8.6)

32. Image classification: example
Heads, portraits or busts of women in profile (13.2)
Women wearing ribbons or flowers in their hair (11.6)
Stylized women, including women depicted in caricature f... (10.3)
Heads, portraits, busts of women not in profile (10.0)
Shadows or silhouettes of women (9.9)
Other flowers including daffodils and irises (8.9)
Other jewelry, including pendants, earrings and brooches (8.5)
Other women including hobos, women holding fans and wome... (8.4)
Groups of females (8.2)
Romans, Greeks or Egyptian women (ancient dress) (8.1)

33. Image classification: example
Slices or quarters of fruits, open fruits (13.0)
Lemons, limes (10.9)
Oranges, tangerines and the like; grapefruit (10.4)
More than one drop including teardrops or raindrops (10.1)
A single drop (including a single raindrop or a single t... (9.6)
Other flowers including daffodils and irises (8.4)
Other leaves (7.9)
A single star with six points (7.8)
Hearts (7.3)
Baby carriages and strollers (7.2)

34. Before/After

35. Before/After

36. Before/After

38. Q.1: where is the image search feature available?
Search area B
Filter area

39. Q1: where is the image search feature available?
Search area B
Filter area

40. Q:2: If you are interested in finding brand that are similar to the image below, which strategy will you select? A
Texture B
Shape

41. Q:2: If you are interested in finding brand that are similar to the image below, which strategy will you select? A
Texture B
Shape

42. Future/past webinars:

43. Global Design Database: webinar
How to perform effective searches in the Global Design Database
July 11 at 4:30pm CET

44. Global Brand Database: webinar
How to perform an effective search in The Global Brand Database
July 18 at 4:30pm CET

45. PATENTSCOPE webinar PATENTSCOPE for experts June 26 at 8:30 am
June 28 at 5:30 pm