English | 简体中文

A collection of ready-to-run Jupyter notebooks for learning and experimenting with the OpenVINO™ Toolkit. The notebooks provide an introduction to OpenVINO basics and teach developers how to leverage our API for optimized deep learning inference.

Notebooks with a button can be run without installing anything. Binder is a free online service with limited resources. For the best performance, please follow the Installation Guide and run the notebooks locally.

Brief tutorials that demonstrate how to use OpenVINO's Python API for inference.

Notebook	Description	Preview
001-hello-world	Classify an image with OpenVINO
002-openvino-api	Learn the OpenVINO Python API
003-hello-segmentation	Semantic segmentation with OpenVINO
004-hello-detection	Text detection with OpenVINO

Tutorials that explain how to optimize and quantize models with OpenVINO tools.

Notebook	Description	Preview
101-tensorflow-to-openvino	Convert TensorFlow models to OpenVINO IR
102-pytorch-onnx-to-openvino	Convert PyTorch models to OpenVINO IR
103-paddle-onnx-to-openvino	Convert PaddlePaddle models to OpenVINO IR
104-model-tools	Download, convert and benchmark models from Open Model Zoo
105-language-quantize-bert	Optimize and quantize a pre-trained BERT model
110-ct-segmentation-quantize	Quantize a kidney segmentation model and show live inference
111-detection-quantization	Quantize an object detection model
112-pytorch-post-training-quantization-nncf	Use Neural Network Compression Framework (NNCF) to quantize PyTorch model in post-training mode (without model fine-tuning)

Demos that demonstrate inference on a particular model.

Notebook	Description	Preview
201-vision-monodepth	Monocular depth estimation with images and video
202-vision-superresolution-image	Upscale raw images with a super resolution model	→
202-vision-superresolution-video	Turn 360p into 1080p video using a super resolution model	→
205-vision-background-removal	Remove and replace the background in an image using salient object detection
206-vision-paddlegan-anime	Turn an image into anime using a GAN	→
207-vision-paddlegan-superresolution	Upscale small images with superresolution using a PaddleGAN model
208-optical-character-recognition	Annotate text on images using text recognition resnet
209-handwritten-ocr	OCR for handwritten simplified Chinese and Japanese	的人不一了是他有为在责新中任自之我们
210-ct-scan-live-inference	Show live inference on segmentation of CT-scan data
211-speech-to-text	Run inference on speech-to-text recognition model
212-onnx-style-transfer	Transform images to five different styles with neural style transfer	→

Tutorials that include code to train neural networks.

Notebook	Description	Preview
301-tensorflow-training-openvino	Train a flower classification model from TensorFlow, then convert to OpenVINO IR
301-tensorflow-training-openvino-pot	Use Post-training Optimization Tool (POT) to quantize the flowers model
302-pytorch-quantization-aware-training	Use Neural Network Compression Framework (NNCF) to quantize PyTorch model
305-tensorflow-quantization-aware-training	Use Neural Network Compression Framework (NNCF) to quantize TensorFlow model

Live inference demos that run on a webcam or video files.

Notebook	Description	Preview
401-object-detection-webcam	Object detection with a webcam or video file
402-pose-estimation-webcam	Human pose estimation with a webcam or video file

The notebooks run almost anywhere — your laptop, a cloud VM, or even a Docker container. The table below lists the supported operating systems and Python versions. Note: Python 3.9 is not supported yet, but it will be soon.

OpenVINO Notebooks require Python and Git. For Python 3.8, C++ is also required. Select a guide for your operating system or environment:

Or, if you have already installed Python, Git and C++, please follow the steps below.

python3 -m venv openvino_env
source openvino_env/bin/activate

python -m venv openvino_env
openvino_env\Scripts\activate

Note: Using the --depth=1 option for git clone reduces download size.

git clone --depth=1 https://github.com/openvinotoolkit/openvino_notebooks.git
cd openvino_notebooks

Upgrade pip to the latest version and install the requirements. Install the openvino_env kernel to create a Jupyter kernel for this virtual environment.

python -m pip install --upgrade pip
pip install -r requirements.txt
python -m ipykernel install --user --name openvino_env

If you wish to launch only one notebook, like the Monodepth notebook, run the command below.

jupyter notebook notebooks/201-vision-monodepth/201-vision-monodepth.ipynb

jupyter lab notebooks

In your browser, select a notebook from the file browser in Jupyter Lab using the left sidebar. Each tutorial is located in a subdirectory within the notebooks directory.

To end your Jupyter session, press Ctrl-c. This will prompt you to Shutdown this Jupyter server (y/[n])? enter y and hit Enter.

To deactivate your virtualenv, simply run deactivate from the terminal window where you activated openvino_env. This will deactivate your environment.

To reactivate your environment, run source openvino_env/bin/activate on Linux or openvino_env\Scripts\activate on Windows, then type jupyter lab or jupyter notebook to launch the notebooks again.

To remove your virtual environment, simply delete the openvino_env directory:

rm -rf openvino_env

rmdir /s openvino_env

jupyter kernelspec remove openvino_env

If these tips do not solve your problem, please open a discussion topic or create an issue!

• To check some common installation problems, run python check_install.py. This script is located in the openvino_notebooks directory. Please run it after activating the openvino_env virtual environment.
• If you get an ImportError, doublecheck that you installed the Jupyter kernel. If necessary, choose the openvino_env kernel from the Kernel->Change Kernel menu) in Jupyter Lab or Jupyter Notebook
• If OpenVINO is installed globally, do not run installation commands in a terminal where setupvars.bat or setupvars.sh are sourced.
• For Windows installation, we recommend using Command Prompt (cmd.exe), not PowerShell.

* Other names and brands may be claimed as the property of others.