StradVision Develops Perception Algorithms To Enable Cars To See And Recognize Objects
Founded in 2014, StradVision develops perception algorithms to enable cars to see and recognize objects inside and outside of the car in a safe and affordable manner. Below is our recent interview with Sunny Lee, COO at StradVision:
Q: Sunny, can you tell us something more about StradVision?
A: By supplying tried and true algorithms to various automotive OEMs and Tier 1 suppliers, our goal is to enable cars to utilize whatever hardware is at their disposal to run very sophisticated deep learning-based perception software. We have over 200 deep learning-related patents that are at the crux of automotive perception software development and are continuously adding more crucial IPs for automation and safety.
Q: How exactly does StradVision’s software work?
A: We developed our own network called SVNet, which is a deep learning-based algorithm that learns and adapts to various objects and situations that can be used inside and outside of the vehicle. It is deliberately designed to require minimum computation and power consumption in order to be deployed onto small and affordable hardware, while executing extremely complex tasks such as detecting all relevant objects on the road in real time – hardware affordability and flexibility is critical to bring such a sophisticated solution to the market and to bring about the advent of the autonomous future as soon as possible.
Q: What makes StradVision’s vision processing software for autonomous vehicles unique?
Q: How safe and accurate is StradVision’s software, compared to other software, like Tesla, Mobileye, etc.?
A: StradVision is all about safety. It is one of the very first deep learning-based software companies that acquired the “Automotive Software Process Improvement & Capability Determination” (ASPICE) CL2 certificate, to ensure safety and accuracy during product development and implementation. Our products have received the Guibao (GB) certification, which is a Chinese government issued certificate for automotive safety – passing this rigorous safety test in China is the result of years of effort to develop the most up-to-date Advanced Driver-Assistance Systems (ADAS).
Our SVNet is a deep learning-based algorithm that has a plethora of edge cases already recognized and studied. Therefore, it is especially strong when it comes to harsh weather conditions such as snow, rain, night, urban, backlight, and more. Also, it can be executed when, say, a stray leaf somehow gets stuck in front of the camera, covering half its field of vision. Computer vision cannot do this, in scenarios when the situation deviates from its predetermined ideal conditions. In addition, the software can be updated post-deployment towards advanced features and enhanced safety.
Again, StradVision’s SVNet is modular and features hardware-specific optimization. Tesla and Mobileye, like the iPhone, requires you to get a monolithic, full-stack solution with hardware and software combined, in one chunk – you cannot modify it, customize it, and even optimize it to your unique needs. StradVision provides the opposite – use whatever hardware you want in whatever unique situations you are in.
Q: When do you anticipate vehicles including StradVision’s technology will start to reach the road?
A: This year is our first rollout of our mass production project. On top of this, there are more than a half-dozen mass production projects from level 2 to level 4 automation in the pipeline each coming year, all over the world, so we know we will grow exponentially. On-road deployment is extremely significant since it will objectively prove SVNet’s reliability and safety and make our algorithms more powerful.
Q: What are StradVision’s plans/goals for the next year?
A: We will stick to our plan and aggressively expand in mature regions such as the United States, EU, and East Asia, as well as developing emerging markets such as India, China, and Southeast Asia. Since we are focusing on mid-range vehicle markets, we are expecting to grow in line with developing regions’ robust growth rate, while advancing our robust solutions and deploying new products for further reliability and safety.