Yesterday’s post was a bit of teaser for the setup of the high-speed video. Now let’s see a couple of examples! (For full effect, I’d recommend playing this on full volume in the background. Rest assured, once we have a highlight reel of lizard runs we’ll definitely be creating a montage of our own)
There’s all sorts of cool stuff going on in this video I had never seen before. For example, notice as the fellow is running into the frame at full tilt he does a little hop and lands with all four feet planted and comes to an almost immediate halt. He then looks left to right to survey his surroundings, and as my big scary hand approaches, turns, pushes off with forelimbs and then generates speed with some big back-leg strides that cause his back to twist with the momentum. He navigates the corners cleanly, but slowly, pausing to look around the corner. Remember though, all of this happens within the span of about a second and a half in real-time.
Now look at this enthusiastic fellow:
He comes barreling in and can’t stop to notice the impending wall. His first crash just turns his head but he sure doesn’t look like he’s attempting to negotiate the turn. At the second crash he crumples and decides a different tact might be best; perhaps climbing the wall and getting out (though it looks more like he’s tap dancing). Again, all of this is happening faster than the eye can really register but at 500 frames per second we’re given a new perspective on these two very different runs.
So now comes the analysis, and this is going to be tricky. Menelia recorded 885 videos and each averages about 2.5 seconds in length. At 500 frames per second that works out to some 1.1 million frames of video to process… Know any good books on tape?
2 thoughts on “And some high-speed videos”
With that much data you could train a neural network to respond to the videos and output a characteristic lizard as well as variations due to environmental factors. See “Wired” June 2016, the end of code.
Yes! We’re thinking of something like this – training the computer to identify the outline of the lizard and the position of the white dots. Fingers crossed it’s going to work! I’ll let you know.