How It Works And Why It Matters
Logo detection is the basis for understanding the presence and visibility of brand logos within a picture or video. Logos can be detected in content posted to social media, or within video footage broadcasted or steamed to any channel or OTT platform. Brand logos are typically placed within a sports context as part of a sponsorship agreement. In order to best understand the value of a sponsorship, it is important to first understand the relative presence or absence of the logo. This is then combined with research outlining the memory and retention for the brand logos within the context of the programming in order to establish the exposure value of the sponsorship.
Our artificial intelligence, or computer vision logo detection solution, identifies where and when a logo is present within an image or video. These are known as detections. The logos are submitted to a database and our solution looks for matches between the video/image and the logo. The level to which artificial intelligence believes it has found a match is referred to as its confidence level. Things such as video quality, noise, lighting and/or logo version can influence the confidence level associated with the match.
We are able to train logos within the system and have them activated within a few days. Additionally, our logo detection abilities extend to the placement of the logo within the context of its surroundings. This provides valuable placement data, which can be used to determine where a brand logo appears. Standard placements are defined for detections and vary by sport. Custom placements are available depending on the level of similarity to already existing trained placements.
This logo detection process can be broken down in various steps. First, the service identifies a logo trained in our system as being present within a video or image. The duration of this logo detection is recorded down to the 1/30th of a second (30 fps). Optionally, the value of that detection is determined based on an underlying CPM value, the number of viewers of the content, and a combination of factors including the size, position on screen, clarity, and clutter of the logo within the identified time frame of the detection. Additionally, identification of the context, or placement of that logo, can be provided. Finally other metrics are offered such as share of voice and breakdowns of logo appearance by size.
Computer vision/logo detection is not a perfect science. The service is offered as a reliable way of automating logo detection as a software solution. Computer vision is only as good as its training and its continual improvement. Our base software provides a way to identify inaccuracies that are then used to improve the computer vision models. We do offer optional services to provide an additional quality assurance layer to supplement computer vision logo detection for even better accuracy and detection issue identification.