0

Hi I am working on an audio classification model. I have implemented in python, now I need to port it to c++.

I do feature extraction in python as it is easier. I save these features, into a file. The feature is represented as a 2D array in numpy.(feature is not an image but one can think of it as as an image)

I would like to load them in to an InferenceEngine::Blob. All the examples I see are loading it from an image file (actual jpeg o gif) using openCV matrix. How can I load a binary array into a blob? Can I load 1D array and specify its shape(width and dimension) as I load it into a blob? If I have an int array of size 9, how can I convert it to 3x3 blob using Vino Api.

Can I / Do I have to load it to openCV matrix from a binary file?(The binary file is not an image) , Can I even do it without data being a an actual image file ?

Thanks

Evren Bingøl
  • 1,306
  • 1
  • 20
  • 32

1 Answers1

1

For the inference of non-image data, you can create the blob using the InferenceEngine::Blob classes. Here is the video on the example implementation of the Inference Engine main function call. https://www.youtube.com/watch?v=JwgPgKClU3w

There is no specific example for an array as an input. Check out the following demos and samples for the supported models such as GNMT, BERT, TDNN, and ESPNet.

Rommel_Intel
  • 1,369
  • 1
  • 4
  • 8