Raspberry Pi/OpenCV/Yolo/Python specialist

We are searching for a motion detection/machine learning specialist who can help us to get a visitor tracker solution working on a Raspberry Pi 3 B+. We hired a developer who tried to get such solution working, but after waiting for a long time, we still haven’t received a solution that works. We need to have a solution that has an accuracy rate of 99% or higher. The current solution is made with mobilenet_ssd (caffemodel), pyimagesearch, python and Yolo. We received information that the solution can not work on the Raspberry Pi 3 B+ because the Pi memory can’t process the recorded video fast enough and therefore object recognition is not working. For this solution, we do not need real-time processing as it is a visitors tracker. So if there is a short delay or time buffer to give the software time to calculate and process recordings, then that’s of course not an issue. As machine learning never can be 100% accurate, we want to have an additional method to double check recordings so that when Yolo, for example, has missed a person that with background subtraction method we still can register a visitor or trigger a double check by Yolo. The scope of this project is to check the solution we have now and to advise us if we need to proceed, redevelop the solution or that the solution can be altered. That is the first project. A second project is to help getting a solution working. More information will be given in a PM