Embedded Computer Vision and AI Design Methodology

Hello everyone,

I have been working on computer vision and artificial intelligence projects for a year which can be considered as new . Throughout this duration all of the projects that I built up has developed with python3.6. Such as object detection using yolov5 combined with dimp18 tracking, object classification etc. However these projects both includes socket communication, external calculations etc. vision is not my only input for operation. When it comes to handle multiple inputs I get stuck on code design. Sometimes communication may become bottleneck for code flow. Is there any design methodology or pattern for these kind of projects or any kind of sources. Because I want to know if I am doing correct or not.

Best Regards

DeepStream is GStreamer based SDK.
There are many public GStreamer plugins for socket communication, for example - RTP and RTSP support , which are extensively verified by many users