Using OpenVINO for problems other than Computer Vision

[Catanzaro__Brian]

Couldn't find another forum for OpenVINO, so I hope this is the best place to ask the question...

I'm interested in Neural Networks to solve a variety of problems that are not Computer Vision (multivariable non-linear optimization, Reinforcement Learning, ...). 

1. Is there another forum for this topic?
2. Can (should) OpenVINO be used for problems other than Computer Vision?
3. Should I just model my system on TensorFlow and/or PyTorch and then figure out how to map it to OpenVINO and Movidius?

[nikos1]

 Hello Brian,

I believe this is the right forum. OpenVINO can be used for problems other than Computer Vision. Having a TensorFlow model is a good start. You could then try the OpenVino model optimizer and convert to IR if all layers are supported. Eventually it will run on CPU and hopefully on Movidius too. Post here if you get any issues or need more information about the process. PyTorch is possible too but you may need to convert to ONNX first and then to OpenVino; I feel Caffe and TensorFlow are easier to work with OpenVino.

Cheers,

nikos

[Hyodo__Katsuya]

@Catanzaro, Brian, nikos I do not know if that will help, but there is a way to convert the model to Caffe or Tensorflow. - For example ModelA(PyTorch) --> ModelB(Tensorflow) ModelA(PyTorch) --> ModelC(Caffe) ModelD(ONNX) --> ModelE(CNTK) ModelD(ONNX) --> ModelB(Tensorflow) ModelB(Tensorflow) --> ModelC(Caffe) ModelC(Caffe) --> ModelB(Tensorflow) - Reference https://github.com/PINTO0309/Keras-OneClassAnomalyDetection#13-model-convert

[nikos1]

Thank you for the reference Katsuya-san. Nice cross-framework solution to convert, visualize and diagnose deep neural network models.