I created an issue before 5 months for which a bug report was filed by Shuba but there hasn't been an update yet. The issue is that I am getting different results in c++ in any device with openvino. Python is working correctly. Old posts and respective files can be found here
With python and tensorflow with local GPU:
With converted IR but with CPU(FP 32):
And with python classification _sample.py with NCS2 (FP 16):
All the three above works as expected. But with C++ classification_sample_async.h build with NCS2(FP16) or GPU or CPU :
Could anyone help.
Intel does not comment on future software release dates. We will make an announcement once it's available for download, please keep an eye on the forum.