Is OpenVINO necessary for accessing Movidius X VPU in a custom hardware?
I would like to understand the work process of the Intel NCS2.
What exactly is happening in my system and what is happening in NCS2 when I run an Inference.
Also can Movidius chip be used for performing pre-processing operations?
Does Movidius have the provision of communicating through SPI interface from any processing unit.
self help documents to understand more is also fine for me.
No, OpenVINO is not needed to access Movidius X but then you would have to implement everything yourself (the whole process of converting a machine learning model into instruction sets for the Movidius, and the transporting and sorting of your data).
Moreover, OpenVINO does not interact with the Movidius X directly, it communicates with hardware such as Neural Compute Stick 2 (MYRIAD plugin) or Intel® Vision Accelerator Design Products (HDDL plugin).
Generally, when an inference is run, the model get translated into instruction sets appropriate to targeted hardware which will later receive your input binary blob.
Whether Movidius X by itself can perform pre-processing is uncertain. Nevertheless, some pre-processing can be included in the model IR.
You can refer to Intel® Movidius™ Myriad™ X Vision Processing Unit Technical Library. However, most of the documents are not open for public viewing.
Thank you, Rizal for the quick response.
I will take sometime to go through the documents you referred. And get back if some more help is needed. Hope you can guide me.
Thank you once again.
Thank you for your question. If you need any additional information from Intel, please submit a new question as Intel is no longer monitoring this thread.