Achieve faster inference speeds with Ultralytics YOLOv8 & Intel’s OpenVINO

Speed and efficiency are critical when deploying artificial intelligence models. Ultralytics is excited to share our latest integration with Intel's OpenVINO™ toolkit, which makes it easier to deploy optimized AI models on Intel hardware. This collaboration with Intel combines the power of Ultralytics YOLOv8 models with the efficiency of Intel's OpenVINO™, delivering up to a 3x speedup on CPUs and enhanced performance across Intel's extensive hardware ecosystem, including integrated GPUs, dedicated GPUs, and VPUs.

What makes this integration special?#

Intel's OpenVINO™ toolkit is designed to maximize AI model performance across Intel hardware. Beyond vision, OpenVINO™ supports a variety of tasks, from language processing to audio analysis. By optimizing YOLOv8 models for OpenVINO™, Ultralytics helps users get faster, more efficient AI inference, whether they're developing applications for video analytics, smart cities, or next-gen retail.

For a detailed walkthrough on how to export and optimize your Ultralytics YOLOv8 model for inference with OpenVINO™, check out our step-by-step tutorial on exporting YOLOv8 to OpenVINO.

A peek into the process#

Imagine being able to export your Ultralytics YOLOv8 models directly into a format that's tailor-made for speed and efficiency. That's precisely what this integration offers. With just a few lines of code, developers can transform their YOLOv8 models into OpenVINO™-compatible versions, ready to take advantage of the hardware acceleration provided by Intel. This makes it practical to run AI applications that were previously limited by computational constraints.

The OpenVINO advantage: Simplifying AI development#

The Ultralytics and Intel integration is a practical step forward in the AI development process. By combining YOLOv8 and OpenVINO™, developers gain an efficient way to use Intel® CPUs, which are central to computing in many fields. This makes AI more accessible and efficient for practical applications.

Using OpenVINO™ optimizes the inference process so Ultralytics YOLOv8 models run efficiently in real-world conditions. This enables rapid deployment of AI solutions across a broad range of devices without costly GPU setups, expanding what is possible in areas such as smart city initiatives and retail customer experiences.

Real-world benefits#

• Enhanced Performance: Experience up to 3x faster AI inference on CPUs, with even greater accelerations across Intel's hardware spectrum.
• Versatile Deployment: Write your application once and deploy it anywhere in Intel's hardware ecosystem, from CPUs to VPUs, without additional changes to the code.
• Ease of Use: Access over 80 tutorials that guide you through optimizing AI models, even if you're not a deep learning expert.

Success stories: Benchmarks that speak volumes#

Ultralytics and Intel have put the integration to the test, benchmarking YOLOv8 models across various Intel hardware platforms. The results show OpenVINO™-optimized models consistently running faster than their counterparts without compromising accuracy, from the Intel Data Center GPU Flex Series to the latest Xeon CPUs.

Fig 1. Inference Time per Image by Model Size and Format (Xeon CPU)

Bringing AI to the real world#

This integration is more than just numbers and benchmarks; it's about helping developers bring AI into real-world applications more easily and efficiently. Whether it's enhancing security systems with faster object detection or creating more engaging retail experiences through intelligent analytics, the Ultralytics YOLOv8 and Intel OpenVINO™ integration helps power a new generation of AI applications.

Get started today!#

To try the integration yourself, export a YOLOv8 model to OpenVINO™ and benchmark it on your own Intel hardware. For more information and a step-by-step guide, visit our OpenVINO Integration Docs page.