Defining “Brandprints” –
Brandprints are shapes, marks, and symbols on consumer products that identify the brand behind the product(s). Common brandprints include logos, product labels and packaging, and trademarked text. They are any visual pattern deliberately designed to help people recognize brands.
In the same way individuals can be ID’d from their fingerprints — and even their unique “faceprint” — brands can be ID’d by their brandprints.
That, actually, is the whole reason logos, packaging and trademarks exist. Without them, Bud Light bottles would be indistinguishable from Miller Lite, Coach handbags would be indistinguishable from Louis Vuitton, and Ford F-150s would be indistinguishable from Dodge Rams (at least to the untrained eye…).
Brandprint detection works just like fingerprint identification.
When an anonymous fingerprint is collected and scanned, the print is uploaded to a web application and compared against millions of known fingerprints in several databases. If there’s a clear-cut match, the system serves the matching print and the personal details of who it belongs to; if there’s a partial hit, an operator is notified and they take a closer look.
With brandprints, Deep.ad is trained to scan visual data to detect pixel patterns that contain brands and/or product marks and text. Any feed — however random or poorly organized — of images and video can be analyzed. From there, we grab the brandprint(s) within the content and compare them against myriad databases of known logos and trademarks. When there’s a match, we surface all of the relevant marketing information about the brand and/or product contained in the video. That includes recognizing relevant pricing and offer information, industry classifications, relevant UPC/GTIN product codes, and more.
This technology can be used to identify when and where your brand appears in public content. But it can also be used to monitor your competition’s advertising and sponsored content activity (online advertising, paid social, commercials, product placements) and related User-Generated Content (UGC).
Brandprint (and Fingerprint) Detection Challenges
In practice, there are two reasons print detection technology fails:
1. Low print quality
With fingerprint detection, partial prints, fingerprint obfuscation/smearing, and fingerprint relevance (age) can stump many systems or lead to false positives. Not a good thing given the stakes.
In brand forensics, many out-of-the-box detection platforms face similar problems. Logos and other brand marks rarely appear in natural video and imagery in an uncluttered, high-resolution manner. Several brands often appear in a single frame — especially with creative assets in the retail and CPG industry. And brandprint size and object occlusion can lead many APIs from overlooking obvious brand intelligence that a person would most likely recognize.
The ability to accurately detect and label small, low-resolution, and occluded brandprints separates great Ad Tech platforms from average platforms and actionable data outputs from details and descriptions devoid of value.
2. Insufficient data to compare (or learn) against
With fingerprint detection, collecting a high-quality print sample is only half the battle. They have no inherent value without a large database of known fingerprints to compare it against in the present or future.
With automated brandprint detection, the curation of the platform’s machine learning training data is critical.
Specialization is everything. Computer vision systems that are taught to detect everything from driving surroundings to weapon detection to in-store grocery purchases lacks the industry nuance needed to produce actionable market research for Ad/MarTech agencies.
That’s why we’ve spent years exclusively training Deep.ad to find and extract brandprints and market information from an array of media channels and formats (TikToks, display ads, email offers, commercials, live streams, etc.).
Taking the Leap to Automated Labeling
We might take fingerprint detection technology for granted today, but it was science fiction as recently as 2000 when the Automated Fingerprint Identification System rolled out.
With the dizzying rise in the popularity of digital video — and the steep decline in the influence of many traditional advertising tactics — the need for modern brand labeling and video monitoring software is urgent both for in-house brand attribution and competitive intelligence.
Most agencies still outsource this work at substantial cost. And the labeled data are prone to human error — an understandable symptom of having operators watch hours of video per day from an unending queue with (sometimes) little more than stopwatches and notepads at their disposal.
In five years, the marketing and data intelligence industries will look back on the manual labeling process with incredulity.
Data labeling BPO takes far too long and provides far too narrow a slice of relevant content in today’s rich media landscape. Agencies can finally start shifting away from collating and labeling visual data by hand — and gain all kinds of cost and reporting efficiencies — with machine learning from Deep.ad.
You can learn more about logo detection and other brandprint identification at www.deep.ad!
