I’m a big SciFi fan.  The Alien and Terminator series are two of my favorites.  I’ve seen Aliens so many times that I can recite any scene.  Try me.

Aliens was released in 1986.  Terminator 2: Judgement Day (my personal fav) was released in 1991.  Teenage me didn’t think much about the reality of Ash, Bishop, or the Terminator himself.  These androids, synthetic humans, and autonomous killer robots were just characters on the screen.  Teenager or not, I knew the difference between fact and fiction.

Fast forward 30 years.  Siri is on my phone and Alexa could be on my kitchen counter (but she’s not.  I’ve seen 2001: A Space Odyssey too!)

Talk of artificial intelligence (AI) and machine learning (ML) is not just the stuff of SciFi movies anymore.  It has become mainstream.  From aviation to healthcare to music, the applications of AI in today’s world abound.  

Wildlife research has had a loving and intimate relationship with technology for decades. It was just a matter of time before the wildlife department and the computer science department started talking to one another.  

We use GPS collars that talk to satellites.  VITs that email us when a fawn is born.  Trail cameras that can send pictures straight to our smartphones.  But none of that stuff works without us in the drivers seat…until now.  

Years from now we can say it all started with trail cameras.  Trail cameras are an incredible tool for wildlife research.  The ability to remotely observe critters is game changing. If those camera traps do what they do best, a researcher can end up with 1,000s of images.  Our own camera traps captured over 200,000 images.  Some studies have resulted in millions.  YAY DATA!   

Then you realize someone needs to look at every one of those photos and identify what’s in them…yay data…:( 

I mean if Siri can tell me where the closest coffee shop is then surely she can help look at some pictures.  Right?

Actually, yes.  Some super-smart people at the University of Wyoming just published a paper detailing machine learning models that accurately classify 2,000 images per minute on a laptop similar to the one I am using right now.  Even more exciting is that they have made this software package available to any user of Program R, a widely used programming language and free software environment for statistical computing.

Training “machine learning models using convolutional neural networks” to recognize elk and mule deer is the bomb.  Amiright?  Seriously, I have no idea what that means.  And while this is surely going to transform how quickly we can analysis and assimilate data, it is still a bit creepy.  

It’s just a matter of time before we ask one of these convolutional neural networks to ID a bobcat and it says “I’m sorry, Dave, I’m afraid I can’t do that.”  

-Jeannine Fleegle
Wildlife Biologist

PGC Deer and Elk Section

 

If you would like to receive email alerts of new blog posts, subscribe here.

And Follow us on Twitter @WTDresearch

Please follow and share:
RSS
Twitter
Visit Us
Follow Me
LinkedIn
Share