Intel's Vision Algorithm Designer Developer Guide

This Developer Guide provides fundamentals of using the Intel's Vision Algorithm Designer.

Intel's Vision Algorithm Designer (VAD) is an IDE on top of OpenVX* that facilitates development of OpenVX algorithms, workloads and capabilities in an intuitive and visual manner. VAD can work with any implementation of the OpenVX 1.1 and OpenVX 1.0.1 specifications by Khronos*.

The OpenVX specification is designed to take the advantage of optimization opportunities such as parallelism, computations offloading, kernels fusion, pipelining, tile processing and more. Computer Vision (CV) applications and algorithms employ Directed Acyclic Graphs (DAG) structures to represent and execute the flow of their algorithms and applications - these structures enable such optimization opportunities.

Running on a heterogeneous hardware environment provides opportunities for performance improvements via scheduling of a node or a sub-graph to the best suited available device, in terms of latency or load balancing.

VAD addresses all of the requirements and features that are needed for the OpenVX user and application developer to build and execute a graph. Such features include but are not limited to: graph creation, visualization, verification and execution, code generation, profiling and performance measurement, debugging and logging, and connecting to remote target devices.

Vision Algorithm Designer Core features

Below are the core features of Vision Algorithm Designer (VAD):

  • Create and modify OpenVX* graphs using a drag-and-drop interface
  • Build, verify and execute OpenVX graphs
  • Generate structured and flexible OpenVX* C code
  • Provide reusable sub-graph representation of complex graphs
  • Save and load the graphs as GraphML files
  • Visualize final and intermediate results of graph execution
  • Show performance measurements of each OpenVX* kernel
  • Utilize OpenVX user and custom kernels
  • Display OpenVX logger output
  • Create plugins to extend capabilities
  • Connect to and control remote instance of VAD
  • Execute graphs in CLI for batch or regression usage
  • Generate a template C code for OpenVX* user kernels.

Typical next step: Getting Started with Intel's Vision Algorithm Designer