openvino | OpenVINO™ Toolkit repository | Machine Learning library

Use InferenceEngine::CNNNetwork::reshape to set new input shapes for your first model that does not have batch dimension.

Up till this moment, there's no official support for OpenVINO integration with .NET application. Instead, the OpenVINO has its own Inference Engine application that supports both C++ and Python. You may refer here for more info.

Performance-wise, since you mention speeding things up, you could try to use OpenVINO Post-Training Optimization Tool to accelerate the inference of deep learning models.

Plus, ensure to choose the right precision for Deep Learning model according to the hardware you are going to use for inferencing.

The linker error shows that it cannot find the TBB symbols. The TBB library should be pointed to by the TBB_DIR variable. You don't have to set those variables manually using cmake's set() function. Instead - in the shell where you compile your own app - you can source OpenVINO's setupvars.sh script. Just run something like: source /opt/intel/openvino_2021/bin/setupvars.sh and re-run the compiler. I can see you're using CLion, not the terminal directly. In that case you can try adding the variable manually. The TBB location might be slightly different between the OV versions but in general it should point to a subdirectory of /opt/intel/openvino_2021 - just browse the installation directory and try to find it or source setupvars.sh in the terminal and copy the TBB_DIR env var value to your IDE.

The optimum way to use this Multi-plugin with multiple devices is by configuring the individual devices and creating the Multi-Device on top.

For example:

The validated and supported Operating System for Windows by Intel® Distribution of OpenVINO™ Toolkit is Windows 10, 64-bit. Using older Windows version might contribute to unexpected issues.

If only looking on your current encountered error, it might be due to Windows PowerShell compatibility version. As it is mentioned in the ffmpeg-download.psl, which requires PowerShell 4+. While Windows 7 installed with the default version, Windows PowerShell 2.0.

The error you encountered: ImportError: libpython3.9.so.1.0: cannot open shared object file: No such file or directory was due to missing external dependency on Heroku.

Follow the steps below to resolve this issue:

Add a runtime.txt to your app’s root directory to specify a Python runtime. Refer to Selecting a runtime.

This issue (running YOLOv4 model with Object Detection Python Demo on MYRIAD device) is a known bug in OpenVINO 2021.4.752. Our developers are fixing it.
On the other hand, I’ve validated the YOLOv4 model using Object Detection C++ Demo and it is working fine. For now, please use Object Detection C++ Demo as an alternative demo.

Generally, Tensorflow is the only network with the shape NHWC while most others use NCHW. Thus, the OpenVINO Inference Engine satisfies the majority of networks and uses the NCHW layout. Model must be converted to NCHW layout in order to work with Inference Engine.

The conversion of the native model format into IR involves the process where the Model Optimizer performs the necessary transformation to convert the shape to the layout required by the Inference Engine (N,C,H,W). Using the --input_shape parameter with the correct input shape of the model should suffice.

Besides, most TensorFlow models are trained with images in RGB order. In this case, inference results using the Inference Engine samples may be incorrect. By default, Inference Engine samples and demos expect input with BGR channels order. If you trained your model to work with RGB order, you need to manually rearrange the default channels order in the sample or demo application or reconvert your model using the Model Optimizer tool with --reverse_input_channels argument.

I suggest you validate this by inferring your model with the Hello Classification Python Sample instead since this is one of the official samples provided to test the model's functionality.

You may refer to this "Intel Math Kernel Library for Deep Neural Network" for deeper explanation regarding the input shape.

I've already shown you how to debug such problems. Well, let's see.

The list of available packages for tensorflow 2.4.1 includes wheels for Python 3.6-3.8. No 3.9 and no source code. Wheels for Python 3.9 are available starting from tensorflow 2.5.0rc0 — exactly like is said in the error message.

What can you do? 1) Downgrade once more, to Python 3.8. Or 2) Use more recent OpenVino source code; the current sources at GitHub list tensorflow~=2.5 as a dependency. Or 3) Find in your downloaded sources files requirements*.txt and replace version tensorflow~=2.4.1 with 2.5.0.