I am encountering a problem performing inferencing on the NCS2 together with the Raspberry Pi 3. I so far installed OpenVino properly, and I am able to run the OpenVino examples without any problem on the Raspberry Pi with NCS2.
However, when I start to run my own model for inferencing, I encounter this problem:
I'm using the C++ sdk for this. So far, I'm not doing anything fancy yet and I'm just modifying from the example code to run my model with my own input images. I trained my model in PyTorch, converted it to ONNX format and ran it through the model optimizer, targeting it for FP16 accuracy as required by the NCS2.
The error seems to be occuring on this line of code:
Initially, I thought that I may corrupted my model file during the conversion process. However, I have tried to run the inferencing using the python SDK, and everything just works fine. I'm kinda in a pinch here, and I don't know what is wrong. Any advice is much welcomed.
An update to the problem,
I have discovered the root cause to the initial problem to be me setting the wrong precision for the input info. Initially I set it to FP16, for the NCS2, when it should be U8. I have changed it to U8 and it manages to run, but now I encounter a segmentation error problem. I found another post here with the similar problem:
I have yet to try out the solution yet. Will update once again when i do.