Image Tech and the Future of Brand Establishment and Protection
There are few phrases more commonly (mis)used in IP and brand protection than “machine learning” and “image technology”. In fact, image tech is a general phrase used to describe several different features which are trained through machine learning strategies to recognize and label objects within any image.
Through Computer Vision (CV), which is field of techniques designed to help computers understand the content of digital images, developers can teach an algorithm what an object looks like so that it can then find and display any images featuring those same objects. What is an ‘object’? It can be anything it’s programmed to be – a shoe, a brand logo, or any other recognizable mark.
So, once the machine has learned to recognize or ‘see’ the bits of data that make up a logo, it can then recognize when it appears in any digital image or video.
This is useful for many different reasons:
- Brand Logo Monitoring for Marketing
- Logo Recognition for Trademark Clearance
- Logo Detection for Brand Protection
- Image Similarity Matching to Stop Copyright Theft
Brand Logo Monitoring for Marketing
Next time you visit your preferred social media platform, have a look at how many of the posts feature images or video. For brand owners who want to know how users are sharing images of their products, and whether they feel positively or negatively about them, logo detection technology can give them insight into the billions of websites, social media posts, video uploads, and images that make up the brand landscape.
On YouTube alone, users upload 720,000 hours of new content every day! Software tools that automatically scan through new videos and identify brand logos are becoming invaluable for companies who want to save time and money.
Logo Recognition for Trademark Clearance and Brand Establishment
In the past, when brand owners or trademark attorneys needed to clear new logos or devices for trademark registration, they relied solely on searches using the text-based Vienna Classification of figurative elements. In order to discover whether their proposed image, graphic label, or other pictorial element was already in use, applicants needed to search for words describing their mark.
Now, however, it’s possible to perform innovative logo detection based on machine learning neural networks. These technologies are rigorously trained using millions of logo image samples and each candidate logo or design is broken down into its component parts and compared with existing logo images around the world. When used in conjunction with the Vienna codes, this delivers greater relevancy and faster results than searching by traditional image codes alone.
Managing brand risk against the threat of infringement and abuse online is one of the key things we do at Corsearch. Logo detection and image tech make that process even more effective because the enforcement success rate increases when images containing logos are found. Via the same machine learning methods, it’s possible to:
- identify the logo(s) to be monitored
- train the machine by identifying images that contain the logo (this involves a one-time effort so the repetitive work is then left to the machine)
- make logos into a search filter which seeks out any online references featuring infringements. This makes enforcement against listings with altered keywords and foreign languages even faster
- perform reverse image searches for listings where the image is the same but the logo is absent. This is often useful when infringers and counterfeiters blur pixels to obscure logos
Image Similarity Matching to Stop Copyright Theft
Original and arresting images don’t just belong to the world’s great photographers, they are heavily invested in by brands of all kinds. Companies produce their own imagery to advertise everything from this season’s must-have handbags to their own corporate initiatives. It’s become too easy though for infringers to steal these images and then repost them on websites, marketplaces, and social media — frequently to advertise fakes, or in ways which infringe trademarks, or cast a brand’s reputation into doubt.
Again, through image tech and machine learning, computers can recognize copyrighted images and then mark them for review based on their similarity with possible copies.
This article was originally published by brandsec’s brand protection partner, Corsearch.
Brandsec is a corporate domain name management and brand protection company that look after many of Australia, New Zealand and Asia’s top publicly listed brands. We provide monitoring and enforcement services, DNS, SSL Management, domain name brokerage and dispute management and brand security consultation services.