Nodar has announced that its 3D perception products, which were previously only available via integrated development kits, are now available as software-only products in a move that “simplifies evaluation and deployment using cameras and compute teams already trust,” according to Nodar founder and CEO Leaf Jiang.
Nodar’s patented stereo vision algorithms are now available through the Nodar SDK as licensable, software-only products, enabling OEMs and developers to evaluate, integrate and deploy long-range 3D perception without dependence on specific hardware.
The stereo vision engine generates dense, real-time 3D point clouds and is designed for autonomy and automation at scale, supporting applications across the automotive industry.
By separating perception software from dedicated hardware, Nodar enables system designers to optimize performance and cost while using existing camera and compute systems.
Products now available as software-only are Hammerhead stereo vision, a depth-perception software that processes stereo image pairs to generate high-density, long-range 3D point clouds with centimeter-level accuracy and reliable performance in harsh environments; and GridDetect 3d object detection, an optional add-on that interprets Hammerhead depth data to detect objects, estimate size, track motion and generate actionable outputs for ADAS and collision avoidance.
The system supports custom stereo and wide-baseline cameras with a simple ZMQ and ROS2 interface, and can be deployed on x86 or ARM platforms with Nvidia GPUs. It enables long-range detection of small obstacles at hundreds of meters and operates reliably in challenging conditions such as dust, fog, rain, glare, vibration and low light. Comprehensive APIs and documentation support faster integration, while use of commodity cameras and compute helps reduce system cost.
“Our goal is to make long-range 3D perception accessible to any autonomy platform,” added Jiang.
Related news, WeRide launches WeChat Mini program for robotaxi service
