OpenVINO_tutorial_cpp

A complete code sample on how to use the OpenVINO toolkit to infer images using pre-trained models in ONNX format, or from the OpenVINO Model Optimizer.

Requirements

• Intel OpenVINO toolkit
• jsonCPP
• Boost

jsonCPP and Boost are not essential for understanding the general idea of the code. jsonCPP is used because the class labels corresponding to the classIDs were in a json file. Boost is used for cross-platform filesystem handling and command line options implementation. I used the OpenCV version that comes packed with the OpenVINO installer.

Environment vars

• BOOST_ROOT points to the boost directory
• BOOST_LIB64 points to the 64 bit Boost libs (most probably $(BOOST_ROOT)\stage\lib)
• OPENCV_411IR_DIR points to the OpenCV dir that ships with OpenVINO toolkit

Usage

ImageInferenceSample --help for help

ImageInferenceSample -m <Path to xml>,<Path to bin> or <Path to onnx> -i <Path to Image Dataset> -j <Path to class labels json file>

Test

Assuming normal build paths (project_dir\release or debug\ImageInferenceSample.exe)

.\ImageInferenceSample -m ..\..\test\ONNX_Models\resnet18v2.onnx -i ..\..\test\ImageNetDataset -j ..\..\test\ImageNetDataset\imagenet1000_clsid_to_human.json

Predicted output should be something like -

-- Dummy image --
Load net time taken = 499 ms
Time taken = 26 ms
Class ID = 398 and confidence = 17.0721
Class name = "abacus"
Image file = "....\test\ImageNetDataset\abacus.jpg"
Time taken = 28 ms
Class ID = 124 and confidence = 16.7009
Class name = "crayfish, crawfish, crawdad, crawdaddy"
Image file = "....\test\ImageNetDataset\bibLobster.jpg"
Time taken = 28 ms
Class ID = 534 and confidence = 14.1346
Class name = "dishwasher, dish washer, dishwashing machine"
Image file = "....\test\ImageNetDataset\dishwasher.jpg"
Time taken = 30 ms
Class ID = 120 and confidence = 16.7068
Class name = "fiddler crab"
Image file = "....\test\ImageNetDataset\fidlercrab.jpg"
Time taken = 29 ms
Class ID = 604 and confidence = 32.6894
Class name = "hourglass"
Image file = "....\test\ImageNetDataset\hourglass.jpg"
Time taken = 27 ms
Class ID = 119 and confidence = 15.943
Class name = "rock crab, Cancer irroratus"
Image file = "....\test\ImageNetDataset\rockcrab.jpg"
Time taken = 30 ms
Class ID = 6 and confidence = 16.713
Class name = "stingray"
Image file = "....\test\ImageNetDataset\stingray.jpg"
Time taken = 27 ms
Class ID = 897 and confidence = 18.8466
Class name = "washer, automatic washer, washing machine"
Image file = "....\test\ImageNetDataset\washingmachine.jpg"
Average time taken = 28 ms