Lynsie Campbell wanted her startup’s logo to represent a bike, but in a figurative way that wouldn’t bash people over the head with the concept.
Yet the sleek, straightforward logo she had labored over kept reminding people of one used by a National Hockey League team — none other than Pittsburgh Penguins nemesis, the Philadelphia Flyers.
That type of issue is a reality for innumerable companies — and a main reason that TrademarkVision, a computer vision and machine learning technology company in Uptown, has entered the business of visual trademark searching.
Ms. Campbell, founder of the new bike lane mapping app, LaneSpotter, said she and her designer hadn’t checked for similar logos during the branding process. “I don’t know of a way to do that honestly,” she said. “You kind of have to hope for the best.”
Enter Sandra Mau, CEO of TrademarkVision, which was founded in Australia in 2011 and also has a space in San Francisco.
Ms. Mau, who went to graduate school at Carnegie Mellon University for robotics, said her company made the move to Pittsburgh to establish a 24/7 development team, stating that she had a great network of colleagues here. The company has 24 employees to manage its three offices, with seven working in Pittsburgh at the StartUptown coworking space.
Read More
Ms. Mau formerly worked as a computer vision researcher for an Australian information and communications technology research center. She designed facial recognition and camera tracking technology that could detect people in airports and shopping malls.
When she left to form TrademarkVision, she applied the same concepts in machine learning and computer vision to trademark protection. Rather than faces, “We decided to train our system to recognize objects,” Ms. Mau said.
“We think of ourselves as the Google of trademark search,” she said.
The U.S. Patent and Trademark Office provides a free tool for identifying logos that may already have been registered as a unique trademark.
That tool operates under a system called “design codes,” said Jacob Rooksby, associate dean of administration and an associate professor at Duquesne University School of Law.
When someone files an application to register a logo as a trademark, the Patent and Trademark Office enters numerical codes that correspond with certain features of the logo, Mr. Rooksby said. That means there are separate codes for designs that have bike wheels and diamonds, for example.
If the system works right, a consulting service that hoped to use a diamond in its logo wouldn’t soar too close to the logo of another company in the same industry who is also using a diamond.
But the Trademark Electronic Search System, or TESS, is pretty counterintuitive for the casual user and looks as if its interface has not been updated in a decade.
“Many [companies] do just ‘hope for the best,’ either because they don’t know how the system works, or they don’t have the money to find out,” Mr. Rooksby said.
Plus, Ms. Mau noted, different countries use different systems. In Australia, the equivalent of U.S. Design Codes are Keywords and in Europe, they’re called Vienna Codes, she said.
TrademarkVision works to boil down the complicated process into a reverse image search. Platform users must simply upload an image of their proposed logo and wait a few moments for a result with similar characteristics.
Ms. Mau demonstrated how many results TrademarkVision’s search tool turns out. For the Mercedes logo, 170,000 results popped up under the U.S. system, along with 50,000 Vienna Code results and 60,000 Australian Keywords.
Just as people have difficulty discerning images, computers can still make mistakes.
To help the algorithm learn, Trademark Vision assigns confidence values to each term that could describe the picture. When Ms. Mau uploaded the logo for Apple Inc., the term “Apple” came up with 83 percent confidence, followed by “one fruit” and several other less-pertinent descriptors.
The Australian government’s trademark office, IP Australia, and the European Union Intellectual Property Office have each officially adopted TrademarkVision’s search engine for trademarks. In the EU model, up to 2,000 searches are conducted daily, on average.
Currently, TrademarkVision offers different subscription service packages, depending on the size of the firm. The company’s search product is software as a service and may also be used to identify other companies who infringe on an existing trademark.
In other words, once a company establishes a trademark, it can monitor other logos for similarities through the TrademarkVision portal.
A single image search costs $400, a text search is $100 and an annual monitor is $600 per year. There are a number of monthly bundled packages. Custom plans are also available.
Ms. Mau declined to disclose financial results.
Often a search that turns up something too similar just leads to a company changing its logo.
Litigation tends to arise when logos are arguably different or industries involved are different, and the outcome of the dispute is unclear, Mr. Rooksby said. He described a case wherein breakfast cereal company Kellogg, which owns trademark rights in a depiction of a toucan, sued a golf equipment company that also used a colorful toucan image in its advertising.
“The court sided with the golf company, reasoning that purchasers of breakfast cereal were not likely to see the toucan on the golf equipment and think that Kellogg had entered that business, or sponsored or affiliated with a company that operates in that business,” he said.
While Ms. Campbell most likely won’t have her app confused with the Philadelphia Flyers logo, she still believes trademark search is key for companies in their early stages of branding.
“Anything creative, yes it is coming from an individual, but a lot of ideas and a lot of designs exist already,” she said. “You’re just revamping it ... for a designer to be able to compare and contrast their logo’s iterations would be so helpful.”
Courtney Linder: clinder@post-gazette.com or 412-263-1707. Twitter: @LinderPG



Comments
Welcome to Post-Gazette subscriber-only commenting community. Please review our community guidelines.