Skip to content

ONNX Parser is a tool that automatically generates openvx inference code (CNN) from onnx binary model files.

License

Notifications You must be signed in to change notification settings

lcskrishna/onnx-parser

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

42 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Build Status

onnx-parser

ONNX-Parser is a C++ Inference Code Generator tool that takes an onnx binary model and generates OpenVX GDF code for quick prototyping and kernel debugging.

The details of various OpenVX Kernels generated are from vx_nn

Note: Currently only the float-32 tensor datatypes are supported. Special layers such as ROI Pooling Layer, Deconvolution are not yet supported.

Supported Models

Network Name Supported
AlexNet Yes
VGG-16 Yes
VGG-19
ResNet-50

Command-line interface

 % onnx_gdf_generator <net.pb> [n c H W]

Here net.pb is the onnx binary models which is a mandatory file. n,c,H,W refers to batch size, number of channels, height and width of an image which are optional parameters.

ONNX Binary models can be found here : models

Pre-requisites

  1. Ubuntu 16.04
  2. CMAKE 2.8 or newer download
  3. Install the protobuf from C++ install instructions.

Build Instructions

  1. After cloning the repository, create a build folder where the executables has to be present.
  2. cmake -DCMAKE_BUILD_TYPE=Release ../onnx-parser
  3. make

Now, the executables are built and present in the build folder.

About

ONNX Parser is a tool that automatically generates openvx inference code (CNN) from onnx binary model files.

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published