Implementing OpenCV
Opened this issue · 3 comments
The raspimjpeg.py script uses the picamera library to capture frames. Would it be possible to implement OpenCV to process the JPEG frames and send them to the Streameye server?
Frankly I have no idea. I believe the JPEG frames are already encoded by the GPU when using the picamera library. But if you think you can increase the performance, please feel free to experiment and prepare a pull request.
@vdb-sander @ccrisan
I have added facial detection support with OpenCV in my fork, but it is very slow because of the lack of hardware acceleration in OpenCV and the way I send the images to STDOUT. I would really like for a proper implementation of this to come to fruition but I have minimal experience with Python.
@owenthewizard you can write it in whatever language/platform you're familiar with. If it's something useful and stable enough, we can add it to the extra folder of streamEye and document it in a wiki page or so. Users can then choose to add it to the pipeline if they need it.
What I insist on though is that we have this work of yours as a separate program/executable, totally separated from raspimjpeg.py or streamEye programs, so that it doesn't affect their performance in any way.