ViSenze Blog

Computer Vision for Intellectual Property Search

Posted by ViSenze 05-April-2018

Picture this: a world where all kinds of IP - ranging from trademarks, patents and logos to historic imagery and microfiche - can be looked up effortlessly without the hassle of codes, queries, or keyword search. Even more, the anxiety of human coding errors can also be eliminated from the process. The good news? This reality in the realm of IP search is not at all far-fetched as AI-powered computer vision technology gains remarkable accuracy with each passing day. 2018 marks a hiking trend in the development of AI, and it is thus become extremely significant to increase one’s understanding and recognition of visual technology as it becomes increasingly relevant to many professions. More specifically, its transformational and empowering effects are not only limited to creative professionals, but also to other professions outside the creative domain such as lawyers and librarians.

Image Recognition for IP

Back in the 90s, search engines first enabled people to search for intellectual properties such as patents in order to quickly affirm and assess their ideas. And while immense amounts of information suddenly became accessible to the public, other fields were, similarly, going through their own metamorphoses. CAD modeling, 3D printed parts, and camera phones are all examples of creative and visually driven technologies that altered the way people created and circulated information. Today, services and applications have also made an appearance to assist in creating multimedia artworks and creating online storefronts such as secure payment gateways and electronic shopping carts, and so on.

 

No doubt, all of this pose as thrilling opportunities for creativity. But on the flip side, the question of IP escalates at the same time. As opportunities for more significant idiosyncrasy and increasing testing, iterating, and opportunities for setting up of businesses (such as on Etsy), the terrain of knowing which rights belongs to who becomes extremely foggy. By the same token, the structure of IP search is challenging and needs to be thoroughly understood. Two such examples are trademark offices and clearance searches with patent. Similarly, offices around the world depend largely on design coding systems - take for example the international Vienna Classification.

Image Recognition for IP 2

Sequentially, these structures depend on human examination to allocate systemized codes that categorize each mark’s figurative elements so that they become searchable via text. However, with human coding brings about the challenge of human error such as bias, judgment and error. Yet, such a scenario presumes that marks are registered. This is a considerable issue because many design marks that are utilized do not have a design coding in the jurisdictions in which they are registered, or many marks are not even registered. The outcome? An immense sea of design marks that lack effortless searchability in any sort of database.

 

With the power of visual technology today, creators of the 21st century ought to have a more sophisticated solution when searching for IP. That’s not to say legacy databases are going to disappear; they’re not - but it is absolutely necessary to start taking steps to make significant improvements in search. IP databases today are also no longer as simple as the past, with much more detailed data such as more feature-based and visual marks for additionally distinctive features.

 

But ultimately, the question is: why comprehend IP via plain language when it can be brought up to the domain of the actual appearance of - for instance - a mark? Today, developed image recognition technology is capable of spotting similar image marks without codes, queries, or keywords. On top of that, utilizing image recognition technology specifically for trademarks also eases the disquietude of human coding errors. Indeed, as AI-powered computer vision technology continues to attain brilliant accuracy, visual search applications for product cataloging and retail has also been rising at an equally swift rate. It’s playing a large role in assisting companies in various industries develop quicker, shorten customers’ time to purchase in the customer journey, and create customer loyalty. On top of that, people are becoming increasingly receptive and congenial to image search everyday.

 

Humans are naturally visual creatures by instinct, and the hunt for fashion and accessories makes no exception. Visual search in the fashion and accessories domain has great potential in building up widespread acceptance and understanding of the applicability of visual search; more often than not, the retail industry faces IP issues as independent designers condemn fast fashion brands such as Zara and H&M for robbing various designs from them. Consider Vetements’ Spring/Summer 2016 collection versus H&M’s (2017 product):  

Intellectual Property Image Recognition

The bottom line is, people have to stop diminishing IP search to mere keyword, code and query search, and taking unnecessary risks due to human error. It is definitely more helpful to familiarize oneself with visual search when it comes to looking up all kinds of IP. Indeed, by partnering both domain-specific expertise with IP and image recognition tech, the full potential of diminishing both cost and peril while increasing efficiency can be fully realised.

 

  Best practices,   Deep learning,   Visual Search

Comment