Cambridge Consultants introduces EnfuseNet AI system for autonomous vehicles; leveraging low-cost sensors and cameras
Cambridge Consultants, part of the Capgemini Group, has introduced EnfuseNet, an Artificial Intelligence (AI) system for autonomous vehicles. EnfuseNet fuses data from low-cost sensors and cameras—hardware costing just tens of dollars—to generate high-resolution depth data, the ideal reference point for autonomous systems.
The result is very low-cost, high-resolution depth data that enables vehicle manufacturers and automotive suppliers to rewrite the economics of vehicle autonomy.
The global market is expected to decline by more than 3% during 2020 as a result of the COVID-19 outbreak , and may take years to recover. High technology costs mean the automotive industry has struggled to introduce advanced driver-assistance systems (ADAS) beyond luxury vehicles and into the mass market. Meanwhile, the ‘arms race’ to rack up millions of driven miles to capture real-world training data favors a small group of early leaders, blocking new entrants.
Against this background, Cambridge Consultants developed EnfuseNet, a low-cost, high-resolution vehicle perception technology. EnfuseNet will help vehicle manufacturers and mobility technology providers to realize a critical element of a self-driving system at a much lower cost, and to deliver autonomy to new and larger segments of the automotive industry.
Building an accurate and detailed depth point cloud—a 3D view around the vehicle—is critical for autonomous decision making. Today’s autonomous vehicles resolve depth data using two-dimensional camera inputs combined with LiDAR or radar. LiDAR remains the most accurate approach but with unit costs for mechanical spinning LiDAR devices in the thousands of dollars, the technology is prohibitively expensive beyond the luxury market. Radar is lower cost but does not provide enough depth points to build a high-resolution image.
EnfuseNet takes data from a standard RGB camera and low-resolution depth sensors, which cost in the tens of dollars per device, and applies a neural network to predict depth at a vastly greater resolution than the original input. Uniquely, this depth information is per image pixel, enabling the system to provide depth data and a confidence prediction for every single object in an image.
EnfuseNet was trained with synthetic data in a virtual learning environment, performing impressively when tested with real-world data. This enables OEMs and automotive suppliers to overcome the time, complexity and cost constraints of collecting real-world data to train their ADAS perception algorithms. Generating high-quality depth point clouds, with confidence down to the pixel level, means that EnfuseNet improves explainability and traceability, reducing the risk of ‘black box’ decision making in a safety-critical application.
The underlying model is based on a completely novel architecture that fuses Convolutional Neural Networks (CNNs), Fully Convolutional Neural Networks (FCNs), pretrained elements, transfer and multi-objective learning and other approaches to optimize depth prediction performance.