Creating An Automatic Highlight Reel For Sporting Events
from the who-needs-human-editors? dept
Weren’t we just saying that humans were being automated out of jobs? Add to the list of jobs at risk the “highlight reel editor”. Here are some details on a system being developed by a grad student to take a sporting event and automatically yank out the most important parts. I could have sworn we wrote about a similar system recently, but can’t find the post. Either way, this article has plenty more details about how such a system works. It looks for certain “visual clues” that something important has happened. For example, when a major play happens, there are often immediate replays in slow motion, as well as text written on the screen (such as an updated score). So, the system will scan for such clues, and then work backwards to pick out the “important” play. The guy working on the software admits that it works much better on sports that play on a monochromatic field, such as soccer, football or hockey. For a game like baseball, he says the system would need to be “tweaked”.
Comments on “Creating An Automatic Highlight Reel For Sporting Events”
Sports as a declining industry
When only the luxury classes could afford sports, knights had their jousting contests in which honour was perhaps more important than winning. In the past century, sports has worked its way down the socioeconomic ladder, as nowadays ghetto kids aspire to become athletes.
The phenomenon of sports hooliganism has not yet become big in North America, but flash mobs might change that. I wonder if hooliganism could evolve into an Al Qaeda like global network, in which stadiums full of tens of thousands can be turned into a tornado of panic through computer-designed panic generators? A laser swipe across the crowd here, a firecracker set off there, simulated blood packets exploding in someone’s shirt. What if hooligans outpower the police and turn a city into a sea of carnage?
Another, simpler way to ID highlights
While visual cues might be an interesting challenge, a simpler way would be to track the average volume and pitch (and pace?)of the announcers voices. Their response clues you to something that happened 15 seconds earlier.