Compound Eye, developer of 3D sensing and perception software, has opened early access to its Vidas Development Kit. Vidas produces dense 3D output and per-pixel semantic classes in a variety of road and terrain environments using only scalable, passive sensors such as vehicle cameras.
Vidas measures the precise depth of environmental perception without lidar or radar. The software enables vehicles to understand their surroundings in high-resolution 3D and in real time. Applications range from dashboard displays to driver assistance and autonomy. The software runs on a standard embedded computer and delivers ten times the resolution of lidar at a scalable cost, the vendor promises. To do this, the software brings together information from stereo parallax, motion parallax and semantic cues in the environment. It also combines classic computer vision and modern machine learning.
Vidas is based on industry-standard hardware and was developed with OEMs and Tier 1 product manufacturers in mind. Cameras can be mounted almost anywhere on the vehicle for easy installation and integration. Vidas is self-calibrating, can be adapted to a wide range of vehicle cameras and works with OEM or third-party software. As the only dense 3D perception technology, Vidas does not require proprietary hardware.
The Vidas DevKit includes two reference cameras, a computing module, a GPS antenna and the necessary cables. The initial installation takes no more than thirty minutes and the results of the acquisition can be displayed immediately on a tablet or laptop.
The Vidas SDK includes Vidas Operator for viewing and recording data and Vidas Inspector for analysing capture results.