I was wondering if the developers or anyone on this forum had any experience with multithreading in regards of processing camera streams for facial recognition or something similar. In particular, if someone could point me towards a code sample or explain how they got their code to work. I am attempting to use multithreading as a way to improve performance in terms of checking against multiple faces but am running into problems. Alternatively, if someone knows how to switch the recognition database without having to rebuild the whole pipeline(SenseManager,FaceConfig,RecognitionConfig, etc), I'd be interested to hear about that as well.
I have worked in computer vision for quite a few years now. There are many steps involved in getting an application like that running correctly.
Here is a video I made to get you started with the initial frame grabbing from the RealSense camera using multi threads:
Image processing itself can be a very time consuming process, so it must be done in a separate thread. This means that you can process an image while you are acquiring the next one from the camera, which may allow you to run in real time.
Thanks! I ended up progressing further in my code but now I have hit another obstactle in terms of hitting a System Access Exception when attempting to create multiple Sense Managers on different threads utilizing the same camera. In fact, I believe it's the almost identical situation the below topic from 2 years ago:
I am attempting to dispose the sense managers correctly but it seems not to be synced just like how the guy in the above thread experienced. I was wondering if anyone has had any experience with syncing Sense Managers from different threads?