Our case is the following problem:
At the moment we use stick v1 with OpenVino.
Our movidius configured to execute predictions at 4fps speed.
When Movidius starts to overheat, predictions slows down depends on the temperature.
Finally, Movidius getting stuck and does not return any predictions.
I succeeded to create the mixed c++ project, using the source of APIv1 for Windows and APIv2 for Linux and now we are able to receive the temperature on Windows. Unfortunately, the Movidius firmware for APIv2, that I am using does not support Stick v2 and is not compatible with OpenVino's firmware. But we can use it now for a while.
However, after the movidius stopped to return predictions, after 45 seconds timeout we reset it and receive its temperature, which is about 73-75 degrees. We don't know exactly, what temperature is at the moment when it stuck, but it's definitely higher. Now, thanks to this function, we are also sure enough, that the problem is in overheating.
We hope there would be some functions in OpenVino to monitor the current Movidius state: temperature, memory consumption and so on. And maybe some internal functions to reinitialize or reset the device during its work.
Unfortunately, it is currently not possible to get the current temperature of the NCS with OpenVINO. Since overheating seems to be the problem here, implementing a small sleep between several inferences could be a possible solution for keeping temperatures low and optimizing predictions. There is a NCSDK sample code (I know you're using OpenVINO but this could be a good starting place) that is able to run for several hours without overheating because it rests for a bit after running several inferences on a video.
You can find that sample here.
Are you getting any other types of error messages? Can I also have more information about your system? Exactly what system are you running your NCS on, etc.
Also, all future OpenVINO questions can be posted here in the Computer Vision forums so that you can get the help you need!
I hope this information was helpful!