This tutorial provides step-by-step procedures for building a simple application that demonstrates some core VPI programming concepts. It reads an input image file, converts it to grayscale, blurs it (all using CUDA backend), and write the output to disk.
| Example input image | Output image, blurred |
|---|---|
 |  |
Officially there are two ways to use VPI. One is via its Python API, the other is via the C API. The former offers a simplified way to use the library that leads to faster development time. Resource management is done automatically under the hood, leaving the programmer with the task of describing the processing pipeline in a simple way. The C API, on the other hand, gives the programmer explicit control of how resources are managed, and is suited for cases where performance and memory usage are at premium.
It is available in the two programming languages officially supported by VPI: