South Korean AI semiconductor company DEEPX is partnering Ultralytics to integrate its hardware into the YOLO computer vision ecosystem. The collaboration aims to establish a hardware standard for Physical AI applications in robotics, autonomous driving, and industrial manufacturing.

The YOLO (You Only Look Once) architecture is used for object detection in factory automation and smart city infrastructure. By making DEEPX a native export target, the companies intend to simplify the process of moving AI models from cloud environments to edge devices.

The partnership embeds DEEPX’s Neural Processing Unit (NPU) software tool chain into the Ultralytics platform. Developers can now deploy computer vision models to DEEPX hardware using a “format=deepx” command. This integration includes automated model optimisation and validation for the DX-M1 NPU.

Physical AI requires high-performance computing with low power consumption to operate autonomously in industrial machinery. The companies have established a continuous integration and deployment pipeline to ensure compatibility between DEEPX hardware and future Ultralytics AI model updates.

“By making DEEPX a native, frictionless export target within the Ultralytics ecosystem, we are defining the new global blueprint for how real-world intelligent systems are built and scaled,” said Lokwon Kim, CEO of DEEPX.