The Value Of 30 Frames Per Second
Video is basically a series of still images viewed in order at a certain speed, to showcase motion. Each of those images is called a “frame.” Most sports and video game content are shot between 30-60 frames per second.
Logo detection within video content relies on breaking the content back down into that series of images and then using computer vision, a form of logo detection, to identify logos and their position on screen, aka placements. There are many ways to approach this service with varying levels of precision and accuracy. Almost all of them require a decision to be made regarding speed, accuracy and ultimately cost. At MVP, we are focused on accuracy and transparency. As such, we choose accuracy over speed. Our computer vision segments all video into 30 frames per second. In other words, every video is split into 30 individual images for every second of video duration. Each of these 30 frames is then scanned for logo detection. This provides the ability to detect logo changes down to 1/30th of a second.
Other solutions will use 1 or 5 frames per second, both to increase the speed of the detection process and lower the cost to scan. This results in rounding to the nearest second or 1/5 of a second, which leads to potential errors, especially when looking at content with a longer duration. Consider a logo that appears repeatedly for a fraction of a second that is rounded up or down. This could result in a significantly greater or a lesser duration of a logo's visibility on screen than reality.
Additionally, sports content moves quickly. A fast-break in the NBA happens in a very short window of time. A shot on goal in hockey happens in a fraction of a second. When a logo detection solution is only looking at 1 or 5 frames per second, it is likely that the chosen frame may contain a blurry image as a result of a motion blur from a camera pan. This means that the logo will not be detected. By splitting every video into 30 frames per second, it is more likely motion blur will not impact logo detection.
Here is a video that showcases the difference in quality between 1, 5 and 30 frames per second.