What to do with a trillion photos on the Internet?
The development of new approaches to finding photos on the Internet is now becoming more than a promising direction:
Just because 100 billion images appear annually.
This contributes to 750 million mobile phones purchased by us, as well as 100 million digital cameras leaving the shelves annually . In the near future, the same Google plans to total 1 trillion images in its index.
')
Companies improve text search ( Picsearch.com , Facesaerch.com ). Others work with visual requests (as a request - an image, not a text) or try to combine text and a picture (Riya, Picollator.ru ).
I work in a company that develops photo search solutions based on image recognition technologies. Of course, I am very interested in new search engines and demos that classify or search for images on the web.
Of the latest releases, the most interesting solution is from Idee Labs . This Canadian company develops systems for classifying and searching for photos.
Idee Inc has shared several photo search demos. Demos work with a closed base, so to say that the company has released a full-featured photo search engine on the Internet is not necessary. It is rather a closed solution for photo stocks, photo catalogs, etc.
The first demo is a search for photos by color.
How it works? The user can choose a combination of colors (up to 10), the demo is looking for photos in its database on which these colors are present.
The second demo allows you to manipulate the base of photos with a volume of 2,823,639 images. This is also a closed base, provided by the photo stock Alamy.
The photos were processed and grouped by similarity. You select a photo from the database, the demo shows you other similar photos.
I would like to draw your attention to the fact that on the site picollator.ru such an opportunity is also available.
When you click on the Find similar link under the photo, you see other photos from the database that look like this one. In this case, the search engine works with sites from the Internet, not with a closed database.
Demo №3 - Search for similar photos in the downloaded image.
I uploaded a photo with the word Untitled, as you can see, the demo gave me different items similar in color and texture.
What does this mean? The demo does not recognize objects inside the photo. Therefore, this solution is unlikely to be suitable for searching photos on the global network, since we are looking for certain items, certain information about someone or something.
The decision of the company Idee Inc will of course occupy its niche.
Who might be interested?
Designers picking up photos on photostock. And in principle, as a marketer, such a system can be useful for me to search for photos for booklets (and for any marketing materials), since it is important to maintain the color gamut.
You can read the full article on my blog.
Just because 100 billion images appear annually.
This contributes to 750 million mobile phones purchased by us, as well as 100 million digital cameras leaving the shelves annually . In the near future, the same Google plans to total 1 trillion images in its index.
')
Companies improve text search ( Picsearch.com , Facesaerch.com ). Others work with visual requests (as a request - an image, not a text) or try to combine text and a picture (Riya, Picollator.ru ).
I work in a company that develops photo search solutions based on image recognition technologies. Of course, I am very interested in new search engines and demos that classify or search for images on the web.
Of the latest releases, the most interesting solution is from Idee Labs . This Canadian company develops systems for classifying and searching for photos.
Idee Inc has shared several photo search demos. Demos work with a closed base, so to say that the company has released a full-featured photo search engine on the Internet is not necessary. It is rather a closed solution for photo stocks, photo catalogs, etc.
The first demo is a search for photos by color.
How it works? The user can choose a combination of colors (up to 10), the demo is looking for photos in its database on which these colors are present.
The second demo allows you to manipulate the base of photos with a volume of 2,823,639 images. This is also a closed base, provided by the photo stock Alamy.
The photos were processed and grouped by similarity. You select a photo from the database, the demo shows you other similar photos.
I would like to draw your attention to the fact that on the site picollator.ru such an opportunity is also available.
When you click on the Find similar link under the photo, you see other photos from the database that look like this one. In this case, the search engine works with sites from the Internet, not with a closed database.
Demo №3 - Search for similar photos in the downloaded image.
I uploaded a photo with the word Untitled, as you can see, the demo gave me different items similar in color and texture.
What does this mean? The demo does not recognize objects inside the photo. Therefore, this solution is unlikely to be suitable for searching photos on the global network, since we are looking for certain items, certain information about someone or something.
The decision of the company Idee Inc will of course occupy its niche.
Who might be interested?
Designers picking up photos on photostock. And in principle, as a marketer, such a system can be useful for me to search for photos for booklets (and for any marketing materials), since it is important to maintain the color gamut.
You can read the full article on my blog.
Source: https://habr.com/ru/post/30261/
All Articles