Machine Vision Algorithms

Highly optimized for the Intel® Movidius™ Myriad™ platform, our vision algorithms allow you to get to market faster.

Machine Learning & DNNs

Deep Learning has the potential to solve some of the most difficult challenges of modern computing but until now has required desktop-class processors, or cloud based approaches. Intel® VPUs offer uniquely high performance in low-level matrix computation kernels, ample on-chip memory flow management, dedicated hardware blocks, and native support for mixed precision computations. Taken all together this modern architecture enables world-class implementation of various Deep Neural Network (DNN) configurations.

3D Depth

Depth algorithms come in many flavors, depending on what sensors are most appropriate for your product. If your device requires object detection collision avoidance, or 3D reconstruction, depth extraction is a commonly required vision algorithm.

To ensure support for the multitude of depth sensor configurations to choose from, we work with several partners in addition to in-house teams to provide a full range of depth solutions. Our stereo depth solution is qualified with multiple reference sensors to support a variety of use cases.

Tracking & Navigation

Tracking features in a complex scene is fundamentally what enables object detection and tracking as well as indoor navigation where there is no GPS signal available. Pose estimation based on visual and inertial inputs is a key enabling technology for robotics, wearable and augmented reality head-mounted displays. For autonomous robots, this can be combined with map construction to enable autonomous behavior such as route planning and collision avoidance. In Intel® VPUs, the sheer number of parallel operations available support robust visual inertial odometry with a low resource and power footprint. Similar techniques are available for object tracking which is useful in robotics and surveillance applications.

Natural User Interface

For applications such as augmented reality head-mounted systems and wearable cameras, we provide algorithms that implement an interface that recognizes hand gestures. Gesture recognition algorithms are critical in gathering user input for use cases that just don't make sense for keyboard or device-based inputs. Multiple solutions are available to best meet the key metrics of latency, ambient light and robustness for different use cases.

More About The SDK

Our Software Development Kit helps you develop applications for Intel's Movidius™ Myriad™ VPU platform using optimized libraries.

Find Out More

About Myriad VPUs

Learn more about our vision processing units designed for power-efficient visual intelligence

Our VPUs

Discover More

Learn more about how to program using our SDK and processing platform

Watch Video

Related Videos