OpenVINO is an open-source software toolkit for optimizing and deploying deep learning models. It enables programmers to develop scalable and efficient AI solutions with relatively few lines of code. It supports several popular model formats[2] and categories, such as large language models, computer vision, and generative AI.
 | |
| Developer(s) | Intel Corporation |
|---|---|
| Initial release | May 16, 2018 |
| Stable release | 2024.5
/ November 2024.[1] |
| Repository | github |
| Written in | C++ |
| Operating system | Cross-platform |
| License | Apache License 2.0 |
| Website | docs |
| As of | November 2024 |
Actively developed by Intel, it prioritizes high-performance inference on Intel hardware but also supports ARM/ARM64 processors[2] and encourages contributors to add new devices to the portfolio.
Based in C++, it offers the following APIs: C/C++, Python, and Node.js (an early preview).
OpenVINO is cross-platform and free for use under Apache License 2.0.[3]
Workflow
editThe simplest OpenVINO usage involves obtaining a model and running it as is. Yet for the best results, a more complete workflow is suggested:[4]
- obtain a model in one of supported frameworks,
- convert the model to OpenVINO IR using the OpenVINO Converter tool,
- optimize the model, using training-time or post-training options provided by OpenVINO's NNCF.
- execute inference, using OpenVINO Runtime by specifying one of several inference modes.
OpenVINO model format
editOpenVINO IR[5] is the default format used to run inference. It is saved as a set of two files, *.bin and *.xml, containing weights and topology, respectively. It is obtained by converting a model from one of the supported frameworks, using the application's API or a dedicated converter.
Models of the supported formats may also be used for inference directly, without prior conversion to OpenVINO IR. Such an approach is more convenient but offers fewer optimization options and lower performance, since the conversion is performed automatically before inference. Some pre-converted models can be found in the Hugging Face repository.[6]
The supported model formats are:[7]
- PyTorch
- TensorFlow
- TensorFlow Lite
- ONNX (including formats that may be serialized to ONNX)
- PaddlePaddle
- JAX/Flax
OS support
editSee also
editReferences
edit- ^ "Release Notes for Intel Distribution of OpenVINO toolkit 2024.4". November 2024.
- ^ a b "OpenVINO Compatibility and Support". OpenVINO Documentation. 24 January 2024.
- ^ "License". OpenVINO repository. 16 October 2018.
- ^ "OpenVINO Workflow". OpenVINO Documentation. 25 April 2024.
- ^ "OpenVINO IR". www.docs.openvino.ai. 2 February 2024.
- ^ "Hugging Face OpenVINO Space". Hugging Face.
- ^ "OpenVINO Model Preparation". OpenVINO Documentation. 24 January 2024.
- ^ "System Requirements". OpenVINO Documentation. February 2024.
- Agrawal, Vasu (2019). Ground Up Design of a Multi-modal Object Detection System (PDF) (MSc). Carnegie Mellon University Pittsburgh, PA. Archived (PDF) from the original on 26 January 2020.
- Driaba, Alexander; Gordeev, Aleksei; Klyachin, Vladimir (2019). "Recognition of Various Objects from a Certain Categorical Set in Real Time Using Deep Convolutional Neural Networks" (PDF). Institute of Mathematics and Informational Technologies Volgograd State University. Archived (PDF) from the original on 26 January 2020. Retrieved 26 January 2020.
{{cite journal}}: Cite journal requires|journal=(help) - Nanjappa, Ashwin (31 May 2019). Caffe2 Quick Start Guide: Modular and scalable deep learning made easy. Packt. pp. 91–98. ISBN 978-1789137750.