About a year and a half ago I saw Jetpac’s video of their Spotter app and I remember thinking at the time that it would be so cool to get this flying on a drone. Some poking got me a little towards that goal but its not until the past week that I sat down and rewrote the code, integrated it with a quad, and did a little demo. The result can be seen in the video.
Unfortunately, it being autumn and this being the UK there was pretty much no window of decent weather to do a nice flight. Instead I had to make do with muddy boots, blustery winds, and frantic umbrella breaks due to rain showers. It was hard to keep the quad under control in the wind and hence the video is not as smooth as I’d like. Once the weather improves and I add a gimbal, things should be much better.
Comments
It won't have the same computational power as the TK1 but It'll be interesting to see to what extend your code could run on Erle-Brain 2. Maybe a simplified version using a lower framerate?
Drop me an e-mail at victor@erlerobot.com and i'll be happy to ship one to you.
@Victor: haven't tried TensorFlow yet but would like to and this would be a good excuse :) Be great to try the brain 2 though I doubt it has the juice to run everything on board (as nice as it is). But could couple it with a Jetson TK1 ?
This is great, thanks Dirk. The new Google's TensorFlow may add some interesting points to this work. Have you consider using it?
I'm also interested to see whether Erle-Brain 2 could do something similar with its onboard camera and the computational resources it has. Would you be willing to give it a try if I ship you one?
Hi Thomas. Thanks, I know, looking into buying one :)
Cool. This is a timely post. We were just discussing how we could get the NVIDIA boards in the air. (http://bit.ly/1GZJgSj)
I respect the great lengths you went through to link-up multiple Odroids. You could have done some serious damage with one of the new boards. :) NVIDIA just released this whitepaper to brag about their deep learning capabilities.
http://developer.download.nvidia.com/embedded/jetson/TX1/docs/jetso...
The punchline is that "the TX1, while consuming only 6W, classifies more images per second with Caffe AlexNet than Intel's high-end i7-6700K (@62.5W)"
Best,
Thomas