High-quality data are the only things that power artificial intelligence algorithms. The continuity of this data flow is essential for algorithms to work comfortably without bottlenecks. Otherwise, the system gets worse and the risk increases. Labeled data is critical for companies that work on autonomous driving, such as Zoox, Cruise and Waymo, who work on machine learning. It is known that Scale AI was established to provide images, LiDAR and maps to companies that create machine learning algorithms. Companies working on autonomous vehicles use the majority Scale, although they use places like Airbnb, Pinterest, and OpenAI to get tagged data. Scale shared a free and licensed data set Due to the COVID-19 outbreak, the vehicles that collect data for autonomous driving have slowed down or have reached the point of stopping. At this point, Scale AI, which wants studies not to slow down, aims to share the data that can be material for the studies free of charge. This week, the company released a set in cooperation with LiDAR manufacturer Hesai. This free, open source data set called PandaSet is a machine learning model for autonomous driving. Suitable for academic and commercial use, the set contains high-resolution data collected by Hesai. The images were collected from the urban areas of San Francisco and Silicon Valley prior to the authorities’ calls to stay at home. The data set contains 48,000 camera views and 16,000 LiDAR scans, according to the company. In an interview with TechCrunch, Sclae’s CEO Alexandr Wang said, “Artificial intelligence and machine learning have great potential, but it is incredibly difficult. Machine learning is a waste unless high quality data is available. That’s why we created Scale and today we aim to take the industry forward with open source code. ” Adding that there is currently the greatest need given high-quality labeling, Wang wants to facilitate access to this data. Of course, all this does not indicate that Scale, a for-profit venture company, will share all the data it has. However, new data with open source is planned to be launched later this year.
