Stradvision (CEO Kim Jun-hwan) announced on the 14th that it will participate in CES 2024, the world’s largest IT and home appliance exhibition held in Las Vegas, USA, in January next year and unveil a new version of SVNet.
Stradvision’s SVNet is an AI (artificial intelligence)-based image recognition software, and an updated version applying the next-generation ‘3D Perception Network’ will be unveiled at this exhibition.
This is a next-generation approach to advanced driver assistance systems (ADAS) and autonomous driving (AD) systems, and expands the role of cameras to support autonomous driving sensing, perception, planning, and action. The company explains that it provides a comprehensive solution that includes 2D to 3D conversion from awareness to planning among the four stages.
Stradvision emphasized that it applied a differentiated strategy in developing the 3D Perception Network. Excluding the LiDAR sensor used in the development stage, it is possible to present an economical and practical alternative to OEM customers by eliminating the need for a LiDAR system in the final vehicle.
This new approach to autonomous driving systems is expected to reduce the overall cost of implementing ADAS and AD technologies in vehicles and entail the expansion of autonomous driving functions throughout the automobile industry. Accordingly, we can expect widespread system adoption and utilization by OEM customers.
According to Stradvision, 3D Perception Network has significantly increased scalability through improved deep learning and reduced post-processing, thereby reducing ARM resource usage by about 30%. As a result, it ensures the flexibility to scale across various SoC platforms and can adapt smoothly to various hardware environments.
In addition, efficiency has been optimized for the SDV (Software Defined Vehicle) era by reducing CPU usage for extensive system operation and increasing NPU utilization for ADAS functions.
Complexity has also been reduced. By implementing the 3D Perception Network, the code complexity of the post-processing process is significantly reduced by up to 50%, and the integration and maintenance process is simplified. At the same time, the innovative approach redefines the role of cameras in the vision area. The role of vision in the autonomous driving system has been expanded by replacing the LiDAR function solely with a camera.
In addition, SVNet provides a more economical option for OEM customers who want to apply cutting-edge ADAS and AD technology to their cars, and the excellent accuracy guaranteed by the 3D Perception Network dramatically reduces detection errors by 10% and improves stability by 30%. It is effective.
StradVision CEO Kim Jun-hwan said, “Our company, which is set to IPO in 2024, has achieved the milestone of cumulative production of 1.5 million units in 4 years since we started producing SVNet commercial products in 2019.” He added, “At this CES, StradVision is leading the automotive sector. “We will once again show how to overcome technological limitations,” he said.
Meanwhile, ‘CES’, the world’s largest IT exhibition hosted by the Consumer Technology Association (CTA), is an exhibition that provides a glimpse into the flow of innovative technologies around the world and is held every January in Las Vegas, USA. AVING News, which has covered CES for 18 consecutive years, has reported CES news around the world every year since 2006, and has been selected through ‘BEST OF CES’ and ‘BEST OF MADE IN KOREA’, which select the best innovative products, companies and technologies. has been in progress. In addition, CES on-site issues are vividly conveyed through ‘AVING LIVE’.