What We Do
|
|
Image Recognition Search: A New Wave of Search Engine Technology
Images on the Internet are growing exponentially.
Keyword image searches account for ten percent of Google’s traffic,
and that traffic volume doubles year-over-year.
Currently, the Internet is not searched using images themselves as a search criterion.
According to IDC white paper, 95% of Digital Universe is not searchable via keywords.
Attrasoft offers search capabilities in two ways:
There are many potential applications for image searches.
For example, corporations can use the accurate identification of an image
to track exclusive images such as logos, trademarks, brands,
and proprietary information such as graphs and charts,
or to research their company logo use by identifying all locations of
that specific image on the Internet.
Companies can also use the AttraSeek technology to increase revenue by
targeting the audience viewing their particular logo.
Government and legal applications of this new technology include searching images from internal databases, such as engineering or architectural designs, documents, maps, tanks, …, or even surveillance images. AttraSeek can also be used to identify a mug shot to prevent multiple instances of an identification card, driver’s license, access card, or passport. Another useful application is searching for images (i.e., special symbols, advertisements, or dignitaries) on foreign language Web sites that have no English keywords, and thus cannot be indexed using a keyword search engine.
In an era when “Google” has become a commonly used verb, it’s possible to search for and find nearly anything online. However, with the volume of information housed on the World Wide Web, it can often be difficult to find exactly what you’re looking for. This has long been the complaint when it comes to the average search engine query for information and, not surprisingly, the same can now be said when searching for images. There is tremendous need for an improved way to search for images and obtain more relevant results.
Today, yielding solid matches from an image search depends on what words are used to tag and describe images on the Internet. This is the heart of the problem, however, because not all images - including graphs and maps - can be accurately described in a few keywords. Also, because assigning keywords to images is labor intensive, many images may not be described fully or correctly, whether intentionally or inadvertently through human error.
Mini-AttraSeek (http://www.attraseek.com/) is an image-recognition search engine that aims to solve this problem by bypassing the use of keywords entirely. Designed for use with a single web site, Mini-AttraSeek offers search capabilities using images, rather than keywords, to query for relevant images. In other words, instead of using descriptive keywords to locate images on a web site, a user simply uploads an image to initiate the search. This capability is enabled by algorithms that create a “numerical fingerprint” of the uploaded image, and compare against fingerprints of images already on the web site.
Attrasoft, a market-proven image and video recognition software company, can provide users with fast, accurate, and scalable image search capability.
Other Applications
Attrasoft detects copyrighted images and videos on the Internet.
Our approach is similar to law enforcement’s use of fingerprinting to find and match identities.
The unique attributes of a video clip – its “video fingerprint” - are compared against a library of video fingerprints that have been taken from the media vault.
The content recognition algorithms that create and match the fingerprints are our strength.
|
|
Who Needs Attrasoft?
Market Pain
- Searching for images and videos with Keywords covers only 5% of the total images and videos (IDC Whitepaper).
- Searching for Information with Keywords sometimes yields too many matches.
- Employees cannot memorize large amounts of images (in the order of 1000); image identification in such cases takes a long time and will have many errors.
- Some images, when described by keywords, will yield millions of results, producing low relevancy.
- Some images cannot be accurately described by keywords.
- Some images can be difficult to describe via keywords; for example, 70,000 car wheel rims styles, 2000 car doors styles, …
- Some images can be mislabeled (human error or intentionally such as drivers licenses).
- Keyword search does not group a specific image well in search results (similar images are separated by many images).
- Automated image search to locate engineering drawings, graphs, architectural designs, maps, document images, logos, etc. are required in many applications.
- Government / Law Firms need an automated search to identify Patent drawings / trademarks and more.
- Assigning keywords to images can be labor intensive and expensive.
- Both image index search and image direct search are required.
- Etc.
Attrasoft ImageTagger is an image-recognition software which tags images automatically,
thus avoiding manual tagging problems: labor intensive, time consuming, and unscalable.
- Mini-AttraSeek is an image-recognition search engine which searches a website via images, not keywords.
- Searching for images with images can yield more relevant results.
- Searching for images with images can be faster.
- Keyword search can be time consuming because you will go through 20 pages of links.
Competitive Advantages
- Fast, Accurate, Scalable to billions of Images.
- AttraSeek fills a technology void.
- Market Proven.
Example
Content producers and content owners need Attrasoft to efficiently:
- Detect piracy
- Monitor compliance with license agreements
- Audit revenue sharing arrangements
- Monetize content
- Monitor and track distribution
In addition, sites that enable Internet users to post image and video content need Attrasoft to :
- Protect their organization from litigation
- Monitor compliance with license agreements
- Be good stewards of copyrighted content in order to gain
further agreements with content owners
|
