Two toy paratroopers stuck to the Aspen Snowmass stake.
When writing image recognition algorithms to recognize the difference between toy paratroopers stuck on the stake and actual snowfall, things can get a little tricky. The last thing you want is an alert to go out saying your resort got 13" of snow because someone duct taped a barbie to the stake (yeah that happened).
Our premier season was interesting. Here's a look back at some of our favorites...
The most typical anomalies are ski patrollers cleaning the stake which should be happening every 24hrs if the stake is getting lots of snow. They're usually only in there for five minutes tops so that isn't too bad to code for, even on the higher resolution video feed cams. The problems come when objects stay there for longer periods of time e.g. snowman on the stake platform, etc.
This typically happens once a day: a member of ski patrol cleans off the stake so it doesn't get buried and we can start an new tally for the 24hr cycle.