If by “easy” you mean someone else already spent 5 years and a nice chunk of cash training a model for it, which you get to use. And if you accept that it will not be accurate across all possible species and environments, only very specific subsets.
You’re right, but you could say the same about the National Park GIS lookup.
We spent the five years training the model. Manually. With Captcha data.
Now we’re teaching it not to run traffic signals, hit motorcycles or busses, or try to drive up stairs.
I’m waiting for the day Google Recaptcha will ask me “is that traffic light red?” and after a couple of seconds “hurry up, I’m approaching the intersection!”
But if my tesla cant go how will hit those last few pedestrian
also fooled by drawings and if it’s shit enough the model will be fooled by text saying “bird”
