- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I recently installed the OpenVINO distribution and all the dependencies on my laptop running Windows10. All the provided demos have been built and run well for MYRIAD target case.
Now, I want to use OpenVINO for a new application that is not an object detection by CNN. I have my Neural Network classification model with signals as input. I was wondering if there is any tutorial (beyond that 30+ demos with pre-trained object detection models) to teach me how to implement OpenVINO for new applications running in Windows10?
Thanks,
Ray
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Ray.
That depends on what exact parts of OpenVINO toolkit you would like to integrate into your application. You can start with this guide - https://docs.openvinotoolkit.org/latest/_docs_IE_DG_Integrate_with_customer_application_new_API.html
Also, please make sure you followed all the steps from installation and initial configuration guide - https://docs.openvinotoolkit.org/latest/_docs_install_guides_installing_openvino_windows.html
Best regards, Max.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thank you Max.
I completed all the steps from installation and initial configuration successfully. I could also run NCS2 for "human pose estimation" demos.
I went through the link you provided, but still I have not found an answer for my question. My deep learning model is not RNN, CNN, or RCNN.
My deep learning model is a simple NN model with two hidden layers. Can I implement OpenVINO for a simple Neural Newtork model? or OpenVINO can only be implemented for R/C-NN models? Thanks.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Ray.
OpenVINO toolkit is based and optimized for Convolutional Neural Networks, so I don't think it works with non-CNN models.
But if your model's format is the same as one of the supported ones (Caffe, TensorFlow, MXNet, Kaldi, ONNX), then you could give it a try. However, we cannot guarantee it will work.
Best regards, Max.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Max,
is there any example showing the integration of OpenVINO into a new user defined application?
This is is useful link that you posted:
https://docs.openvinotoolkit.org/latest/_docs_IE_DG_Integrate_with_customer_application_new_API.html
is there any tutorials or demos showing the integration process?
Thanks,
Reza
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Max,
is there any demos showing the integration of OpenVINO toolkit for a user defined application?
Thanks,
Reza
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I used the optimizer and could generate the IR (*.xml and *.bin) files of my CNN model. Could you please let me know what is the next step? how can I implement OpenVINO for deep learning inference on the edge?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi, Morsali, Reza.
With regards to Inference Engine utilization in user applications - in addition to the link that I shared before, please also check these resources:
Inference Engine Samples - https://docs.openvinotoolkit.org/latest/_docs_IE_DG_Samples_Overview.html
Inference Engine Demos - https://docs.openvinotoolkit.org/latest/_demos_README.html
These links provide you with code samples of how IE capabilities might be used further for your specific purposes.
You could also find some of samples and demos being already installed along with the latest OpenVINO toolkit 2020.1 build.
Thanks.
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page