I have used NCSDK version 1 and 2 and noticed that the way of coding, API wise, is very different from each other. I want to know whether there is any standard or specification is been followed or it's just coded as per the improvement.
The Neural Compute SDK v1 and v2 are based on different versions of the Neural Compute API. The NCSDK v2 uses the NCAPI v2 which is not backwards-compatible, applications developed with the NCAPI v1 will not compile or run with NCAPI v2.
Take a look at the Neural Compute API Overview page for additional information. Hope this helps!