Toyota Backs Tesla's Camera-First Computer Vision Approach to Autonomous Vehicles
Toyota will work With Tesla
A recent report indicates that Toyota is siding with Tesla on the future of autonomous vehicles. Tesla is using a camera-first computer vision approach over the more popular LiDAR sensor approach to self-driving cars.
A recent report states that Toyota subsidiary Woven Planet will utilize a camera-only approach for its self-driving project development, joining Tesla as the only other company to adopt a vision-based strategy to advance itself in the race to autonomous driving.
The two companies intend to form a specialist team to further those efforts. TMC has agreed to purchase $50 million of Tesla’s common stock issued in a private placement to close immediately subsequent to the closing of Tesla’s currently planned initial public offering.
"I’ve felt an infinite possibility about Tesla’s technology and its dedication to monozukuri (Toyota’s approach to manufacturing)," said TMC President Akio Toyoda
Tesla does not use lidars and high-definition maps in its self-driving stack. “Everything that happens, happens for the first time, in the car, based on the videos from the eight cameras that surround the car,” Karpathy has said.
Even as Google’s Waymo, GM’s Cruise and many others stake their claim to the future of robo-taxis, Tesla FSD kit remains an uncomfortable under-dog in the race to fully AVs. The technology has taken years longer than many of these companies promised or suspected.
Yet to see Toyota side with Tesla on this debate really is quite something. Toyota seems to understand how LiDAR is too costly and agrees with Tesla on the scalability of computer vision instead. Woven Planet told Reuters it is able to use cameras to collect data and effectively train its self-driving system, a "breakthrough" that it hopes will help drive down costs and scale up the technology.
Cheaper cameras is a lot easier than trying to make LiDAR sensors cheaper. So computer vision is potentially more scalable. Woven Planet’s cameras will be 90 percent less expensive than the sensors it previously used, and installation is easy, which could make scaling the project relatively easy.
Tesla has always taken the approach that cameras are essentially the camera’s eyes, and radar or LiDAR equipment is not necessarily crucial in developing a successful self-driving system. Musk once said LiDAR was a “crutch” and last year, during an Earnings Call, said Tesla would benefit from a camera-only approach for self-driving.
Maybe Elon Musk calling LiDAR foolhardy is justified, Toyota seems to agree. Toyota is the best car manufacturer in the world, even as VM and GM and others invest heavily in AVs, Toyota and Tesla may get their cheaper.
Sublink is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.