I’m currently building an intelligent speedometer which integrates the speed limit into the drivers display. I set up the speedometer interface with Skia with a changing color depending on the actual speed (input via the mouse y-axis). Now, I want to implement a speed sign recognition with live video input (webcam). I’m releatively new to vvvv and coding but I guess it should be somehow possible with OpenCV maybe?
I want to be able to show a speedsign to the webcam and then the system should recognize it and set the speed limit accordingly.
I searched the forum and found this thread from 2013: Beta Text Recognition but I’m not really familiar with beta and don’t know how to translate this to gamma or even know if this is the best way to do this. Has someone already done something like this in gamma and can give me a hint?
Hey bjoern, thank you for the hint but I had a look into Wekinator and discovered that the object detection via webcam as an input source for Wekinator is described as beeing very slow and not yet available for the Windows platform. Since vvvv only runs on Windows at the moment and Wekinator has to run for this to work, it’s not an option at the moment…
If you know in advance where the thing you want to recognize is on your video feed, you could downscale that region and send greyscale pixel values to Wekintor to train and recgonize your signs, as suggested in this Youtube tutorial. Then simply install the VL.Wekinator nuget and have a look at the help patches
Now if you want to do detection, you could train your own model in RunwayML as documented here, either run it in their cloud thing or on your LAN on a Linux machine and then simply query it using VL.RunwayML
For informations on how to use VL.RunwayML and VL.Wekinator, there are a bunch of resources you use :