DeepScale
DeepScale developed efficient computer-vision software for vehicles before its acquisition by Tesla.
Founded
2015DeepScale developed efficient computer-vision software for vehicles before its acquisition by Tesla.
DeepScale's technical approach sought to reduce the computing resources needed for neural-network perception. Its work targeted automotive-grade processors and sensors, broadening the hardware on which vehicle-recognition software could run.
The company described supporting different processors rather than relying on one hardware supplier's neural-network libraries. That emphasis could help vehicle manufacturers integrate perception into varied hardware designs, although the acquisition report did not supply independent performance benchmarks.
Source: techcrunch.com
Vehicle perception software must operate as part of a larger system that includes sensors, mapping, planning, and control. Efficient recognition alone does not establish that a complete vehicle can drive safely without human supervision.
DeepScale's acquisition added technology and engineers to Tesla's Autopilot team, but the reported vehicles remained driver-assistance systems at the time. The transaction also ended DeepScale's independent role as a standalone perception-software supplier.
Source: techcrunch.com
DeepScale developed perception software for vehicle driver-assistance and autonomous-driving systems. Its neural-network approach emphasized running visual recognition efficiently on automotive sensors and processors with limited power and cost budgets.
The software was intended to help vehicles classify their surroundings in real time and was developed for use across different processor suppliers. Tesla acquired the company and its team, so these capabilities describe DeepScale's historical standalone offering.
Source: techcrunch.com