Tuesday, February 14, 2017

Vehicle Detection for Cyclists Using a Raspberry Pi 3 with OpenCV Haar Cascades


Real-time video image detection running on a Raspberry Pi 3 standalone

The results of experiment 4A were a success. 4A, versioned "WOFDAV Alpha POC 0.1", was intended to establish that the hardware, software, and approach was sound, to learn more about how to do this, to find limitations to overcome in future experiments, and to continue the fun engineering process.

I will post further details in future posts. This is just a quick progress update before I go hit the road to get more data.

Better shot of the Pi, with the power bank that puts out 2.4 amps at 5v

WOFDAV stands for Watch Out For Distracted Autonomous Vehicles.

Up next in experiment 4B:
  • Improve memory footprint and performance
  • Work on multi-threading
  • Look into leaner modules
  • Integrate camera
  • Road training
  • Road testing
  • Detection noise filtering (detect unicorns not herds)
  • Experiment 5 is planned to included TensorFlow

Current status: giddy.

 

Monday, February 6, 2017

Deep Bike Dreams


Yes, still just goofing around with neural network image processing, with bikes in it

I have throughout the lifetime of this blog sometimes run in the house with an idea in my head that I just had to type out as fast as possible because it was burning brightly and had to get out. Other times, I have had thoughts stew for the entire ride, mulling around and around, and then whisper out of my fingers into cyberspace. Neither of those happened today. Today I just went for a quick and excellent ride on a warm, Phoenix afternoon, took a photo of my bike against a wall, came home, and ran the photo through DeepDream. I like the way it all turned out.

I promise to write some code, hook up some hardware, and do something kind of useful, soon.