Logo detection
Logo detection is a computer vision task that finds brand logos in images or video frames and reports which brand appears, where it is on screen and for how long. Models are trained on examples of known marks so they can spot them on shirts, signs, products or screen overlays. The output is usually a list of brands with timestamps and confidence scores.
Why it matters
Sponsorship deals are priced on exposure, and logo detection measures that exposure from the footage itself instead of estimates. It also helps with compliance, revealing undisclosed paid placements and competitor logos that might upset a sponsor or advertiser.
How it works
- A model is trained on many example images of the logos to be found.
- Frames are sampled from the video and scanned for those marks.
- Each match gets a brand name, a position on screen and a confidence score.
- Matches are combined over time to report how long and how prominently each brand appeared.
How is sponsor screen time measured in sports video?
Sponsor screen time is measured by running logo detection across the broadcast and adding up how long each logo is visible. Better reports also weigh size, position, clarity and whether the logo was partly hidden. For example, a logo filling a quarter of the screen for ten seconds counts for more than a tiny corner logo shown for the same time.
Can logo detection find new or unknown logos?
Usually not on its own. Most logo detection models only recognize marks they were trained on, so a new brand needs training examples first. Some systems can flag that a logo-like graphic is present without naming it, which a person can then label. Vision language models can also read brand names from text on screen.
Where it's used
Sports and news broadcasters, creator platforms and marketing analytics products use logo detection to confirm sponsor visibility, spot undisclosed sponsorships and measure how long brands appear on screen.
