Parallella Community

Real-time image processing demo?

OpenCV (all code freely available) implements a number of image processing algorithms that have become quite standard now. Take for example, corner detection. This runs as a filter on the acquired image and is highly parallel-izable. So far people have implemented Verilog versions of corner detection on FPGAs (from Xilinx). Given the promise of Parallela it would seem that C/C++ code from OpenCV can be adapted quickly to run on this.

If this can be demonstrated, next steps could happen fast. E.g. complete stereo vision processing pipeline could be implemented on multiple Parellela boards.

Does anyone know of such an effort? Is this possible on Parallela boards?

i dont know of any efforts to do so but i assume it wont be a giant leap to port the OpenCV's GPU code to Epiphany since it's based on streaming data. documentation here: http://docs.opencv.org/modules/gpu/doc/gpu.html