Waymo will share his data of self-driving, but it's still not enough



[ad_1]

Waymo, the standalone product of Alphabet, is the latest company to offer some of the information gleaned in its vehicles to the broader research community.

The news: Waymo said he would share some of the data collected from his vehicles free of charge so that other researchers working on autonomous driving could use them. Waymo is not the first to do it: Lyft, Argo AI and other companies have already open-source some data sets. But the Waymo movement is remarkable because its the vehicles have already traveled millions of miles on roads.

Why is it important? Unlike human drivers, autonomous vehicles do not have an instinctive understanding of the world. Instead, they rely on training data to learn about the conditions they may encounter and how to respond. The higher the IA quality data models need to train, the better.

The Waymo data: It contains 1000 segments, each capturing 20 seconds of continuous driving. Data comes from four locations: San Francisco and Mountain View California; Phoenix in Arizona (where Waymo launched a small-scale robotaxi service) and Kirkland in Washington. It also comes from multiple sources, including cameras, as well as radars and lidars, which reflect lasers on nearby objects to create 3D maps of their surroundings. Fortunately, the company has labeled items such as pedestrians, bikes and signals in the dataset, which means that other researchers will not have to do this tedious job.

Data cluster: While Waymo deserves some credit for his move, he only shares a tiny piece of the information he has gathered. Other companies also accumulate data for reasons of competition and particularly reluctant to share information on accidents and near misses. But if the industry wants to overcome concerns about the safety of autonomous vehicles, the companies that work there will have to be much more transparent about what they have learned.

[ad_2]

Source link