Our Stack

SLAM PIPELINE

Our SLAM Pipeline is a collection of three machine learning models. First model is our Aerial 2 Ground data fusion model. Second model is our Structure from Motion SLAM pipeline. Third model focuses on converting fisheye imagery to dense point cloud information.

READ MORE

REAL TIME MAP GENERATION

Our Real Time HD Map is focused on converting the car’s sensor data into a real time bird’s eye view representation of the scene along with road furniture feature extraction to enable the autonomous car to achieve motion planning.

READ MORE

A Real Time Scene Contextualization Engine

Directly translate the vehicle sensor data into scene context

JOIN WAITLIST
360 CAMERA PERSPECTIVE

The fisheye camera stitching allows for 360° Horizontal Field of View and 180° Vertical Field of View. Training ML models to handle heavy distortion avoids the need for unnecessary dewarping or cubemap calculations.

80 OBJECT CLASSES

The Object Detection and Tracker pipeline seamlessly handles occlusions, arrivals and departures from the scene. The 360° perspective also improves the overall stability of the system when handling large occlusions.

HARDWARE ACCELERATION

The Edge hardware is optimized to handle large camera pipelines with low latency over long distances. Hyperspec AI has developed proprietary technology to get the data from the sensor into the GPU memory efficiently.

SCENE CONTEXT

Replace the High Definition Map in the car’s stack with a real time 3D model with semantic context. The resolution and fidelity of the real time map matches the high definition map schema for seamless interoperability.

How to Engage Us

Car manufacturers are looking to expand the coverage of the serviceable area for autonomy. They need technology to accelerate the expansion of their map coverage. Hyperspec AI helps car manfacturers gain an unfair advantage in obtaining map coverage for self driving and driver assistance applications.

Contact us