Hello NV12 Input Classification C Sample

This sample demonstrates how to execute an inference of image classification networks such as AlexNet with images in NV12 color format using Synchronous Inference Request API.

Hello NV12 Input Classification C Sample demonstrates how to use the NV12 automatic input pre-processing API of the Inference Engine in your applications:




Blob Operations


Create a NV12 blob

Input in N12 color format


Change the color format of the input data

Basic Inference Engine API is described in Hello Classification C sample.



Validated Models


Model Format

OpenVINO Intermediate Representation (*.xml + *.bin), ONNX (*.onnx)

Validated images

An uncompressed image in the NV12 color format - *.yuv

Supported devices


Other language realization


How It Works

Upon the start-up, the sample application reads command-line parameters, loads specified network and an image in the NV12 color format to the Inference Engine plugin. Then, the sample creates a synchronous inference request object. When inference is done, the application outputs data to the standard output stream.

For more information, refer to the explicit description of Integration Steps in the Integrate OpenVINO Runtime with Your Application guide.


To build the sample, use the instructions available in the Build the Sample Applications section in OpenVINO Toolkit Samples.


To run the sample, you need to specify a model and an image:

The sample accepts an uncompressed image in the NV12 color format. To run the sample, you need to convert your BGR / RGB image to NV12. To do this, use one of the widely available tools such as FFmpeg or GStreamer. The following command shows how to convert an ordinary image into an uncompressed NV12 image, using FFmpeg:

ffmpeg -i cat.jpg -pix_fmt nv12 cat.yuv


  • Since the sample reads raw image files, a correct image size along with the image path should be provided. The sample expects the logical size of the image, not the buffer size. For example, for 640x480 BGR / RGB image, the corresponding NV12 logical image size is also 640x480, whereas the buffer size is 640x720.

  • By default, this sample expects that network input has BGR order of channels. If you trained your model to work with the RGB order, you need to reconvert your model, using Model Optimizer with --reverse_input_channels argument specified. For more information about the argument, refer to the When to Reverse Input Channels section of Embedding Preprocessing Computation.

  • Before running the sample with a trained model, make sure that the model is converted to the OpenVINO Intermediate Representation (OpenVINO IR) format (*.xml + *.bin) by using Model Optimizer.

  • The sample accepts models in the ONNX format (.onnx) that do not require preprocessing.


  1. Download a pre-trained model, using Model Downloader :

    python <path_to_omz_tools>/downloader.py --name alexnet
  2. If a model is not in the OpenVINO IR or ONNX format. You can do this using the model converter script:

    python <path_to_omz_tools>/converter.py --name alexnet
  3. Perform inference of the NV12 image, using the alexnet model on a CPU, for example:

    <path_to_sample>/hello_nv12_input_classification_c <path_to_model>/alexnet.xml <path_to_image>/cat.yuv 300x300 CPU

Sample Output

The application outputs top-10 inference results.

Top 10 results:

Image ./cat.yuv

classid probability
------- -----------
435       0.091733
876       0.081725
999       0.069305
587       0.043726
666       0.038957
419       0.032892
285       0.030309
700       0.029941
696       0.021628
855       0.020339

This sample is an API example. Use the dedicated `benchmark_app` tool for any performance measurements.