StradVision, whose AI-based camera perception software is a leading innovator in Advanced Driver Assistance Systems (ADAS) and Autonomous Vehicles, has launched a cloud-based Auto Labeling Tool (ALT) that works with the company’s SVNet solution to quickly and accurately identify potentially hazardous objects and road conditions.
Designed to be used with SVNet, ALT promises to usher in a new era for data labeling, leaving behind many of the drawbacks and risks of conventional manual data labeling solutions.
ALT takes advantage of the software’s deep learning-based embedded perception algorithms to allow vehicles to detect and recognize objects on the roads, such as other vehicles, lanes, pedestrians, animals, free space, traffic signs, and lights, even in harsh weather conditions or poor lighting.
Compared with competitors, SVNet is compact, requires dramatically less memory capacity to run, and consumes less electricity. It can also be customized for any hardware system thanks to StradVision’s patented and cutting-edge Deep Neural Network enabled software.
To achieve surround vision, SVNet’s camera and deep learning-based capabilities work seamlessly with other sensors such as LiDAR and RADAR to process collected road data with high speed and accuracy.
Data crunching for ADAS and autonomous vehicles
For Artificial Intelligence (AI) to be effective in object detection and recognition, the task of gathering and processing data is as important as developing the software.
To bring top-level perception AI to market requires significant investment – but so does data acquisition and labeling. Detecting and tagging data samples involves a detailed training process for all machine learning software, often requiring expensive and time-consuming manual input from human data labelers.
A high level of human contribution also raises the risk of human error, particularly as ADAS data labeling is a repetitive but detail-oriented task.
StradVision, however, solves this problem for ADAS and autonomous vehicles with ALT and the patented deep-learning SVNet.
A game-changing solution from StradVision
StradVision’s ALT system connects the dots between data recorded by a vehicle and its AI software.
For each separate frame of image or video recorded by the vehicle’s ADAS, every object in the frame is labeled by ALT and sorted into three categories: objects (ex. vehicles, pedestrians, traffic lights, road signs, and other static objects), lanes (ex. lanes of traffic in the immediate surroundings) and segmentation (ex. road surface or free road space).
These labeled data points are in turn compiled into a knowledge bank of information as the vehicle’s AI system continues to learn.
Cloud-based speed and accuracy
Using SVNet as a template, ALT can drastically scale-up data labeling and AI optimization through 24/7 data processing with its Graphics Processing Unit (GPU).
Once ALT is deployed, it automatically annotates and labels 97% of objects at eight times the speed of a human being and at a fraction of the cost – removing the need for a large team to spend hundreds of hours correcting objects identified by a vehicle’s AI system.
Where human intervention is required, a small team of experts can make adjustments to guarantee data quality and reduce any potential incidents.
Augmenting the AI capabilities of Tier 1 and OEM partners’ offerings
StradVision is pleased to offer ALT to its automotive Tier 1 and OEM partners, so they can fully utilize SVNet in-house with their own data.
Automakers can use ALT with SVNet to expedite the development and deployment of their ADAS and autonomous vehicles economically, swiftly, and securely.
StradVision’s software has obtained China’s Guobiao certification and the coveted ASPICE CL2 (Automotive Software Performance Improvement and Capability Determination Containment Level 2) certification. It is being deployed in 9 million vehicles – such as SUVs, sedans, trucks and self-driving buses – worldwide in partnership with five of the world’s top auto OEMs. StradVision’s global partners also include NVIDIA, Hyundai, LG Electronics, Texas Instruments, Renesas, and Aisin Group.