Waymo is Sharing its Massive Self-Driving Dataset With Researchers
Before self-driving cars can safely navigate urban streets, the vehicles need large datasets to train the machine learning models that are used for navigation, recognizing street signs, pedestrians and other vehicles. However, training machine learning models requires an enormous amount of data and collecting it is long and painstaking process, especially for many of the budding startups working on autonomous driving.
Today Waymo, which spun out of Google's self-driving car project in 2016, announced that its releasing the ‘Waymo Open Dataset' for researchers and developers working on autonomous driving and other related mobility projects. Waymo says its dataset is the the largest, richest, and most diverse self-driving dataset ever released for research.
The data was collected by a fleet of Waymo self-driving vehicles that traveled over 10 million miles in 25 different cities.
The dataset includes high-resolution sensor data covering a wide variety of environments, include dense urban areas and suburban streets. That data was also collected in a wide variety of real-world conditions, including day and night, at dawn and dusk, in bright sunlight and rain.
This data is an invaluable tool for other parties working on autonomous driving. Waymo's own engineers use the dataset to develop self-driving technology and innovative machine learning models and algorithms. With the release of dataset, engineers outside of Waymo are getting access to the same data the Waymo's uses for the first time ever.
All of this data is then fed into Waymo's Open Dataset then crunched and processed by machine learning algorithms. Developers can also use datasets to improve upon existing algorithms by analyzing the data and using it to improve software to behave more like a human driver.
Waymo's Dataset Will Help Others to Improve Self-Driving Technology
Waymo believes that offering the dataset it will help speed up the development of self-driving technology by sharing data and thereby promoting collaboration among developers, even if they are outside of the company.
"The more smart brains you can get working on the problem, whether inside or outside the company, the better," says Waymo principal scientist Drago Anguelov in a statement.
All of the data has been labeled and formatted to aid in research.
The dataset includes Camera-lidar synchronization. Waymo is currently working on 3D perception models that fuse data from multiple cameras and lidar.
Waymo's dataset contains data from 1,000 driving segments. Each segment captures 20 seconds of continuous driving, corresponding to 200,000 frames at 10 Hz per sensor, according to Waymo. This longer footage allows researchers to develop models for predicting the behavior of other road users.
Each segment contains sensor data gathered from five high-resolution Waymo lidars and five front-and-side-facing HD cameras. The dataset includes lidar frames and images with vehicles, pedestrians, cyclists, and signage that's been carefully labeled, capturing a total of 12 million 3D labels and 1.2 million 2D labels.
The Waymo Open Dataset has the potential to help researchers make advances in 2D and 3D perception, and improve behavior prediction, which is especially helpful for navigating dense urban areas where there are many pedestrians for a self-driving car to deal with.
The dataset includes camera footage from Waymo's high-definition cameras and 1.2 million 2D labels.
By releasing the dataset, Waymo hopes that the research community can use it to make self-driving vehicles more capable and safer. However, the comprehensive dataset can be used for applications outside of autonomous driving such as the related fields of computer vision and robotics.
Waymo's vehicle collected the data in Phoenix, AZ, Kirkland, WA, Mountain View, CA and San Francisco, CA capturing a wide spectrum of driving conditions (day and night, dawn and dusk, sun and rain).
Waymo is not the only company sharing its data for autonomous driving. Researchers at the University of California, Berkeley shared its DeepDrive dataset, which was once the largest dataset for self-driving AI. DeepDrive contains over 100,000 videos of over 1,100-hour driving events across different times of the day, and varying weather conditions.
China's Baidu released its ApolloScape dataset in March 2018 as part of its open Apollo autonomous driving platform. Baidu's ApolloScape dataset has 26 pre-defined semantic items, like cars, buildings, people walking on the sidewalk, traffic lights, street lights, etc. This has been done using a pixel-by-pixel semantic segmentation technique, according to Baidu.
Waymo said that the release of its self-driving dataset is just the first step. The Alphabet subsidiary is welcoming feedback from the developer community on how to make its dataset even more useful with future updates.
The dataset is available free of charge to researchers at waymo.com/open.
- AutoX to Use the 'Eyeonic Vision Sensor' from California-based SiLC Technologies for its Robotaxi Fleet in China
- Tesla Challenger NIO Inc. Delivers its 200,000 Electric Vehicle, a Significant Milestone
- Elon Musk’s Deal to Buy Twitter for $44B is Put ‘On Hold’ as Tesla Loses $400B in Market Cap Since the Plans Were First Made Public
- EV Startup Fisker Inc to Produce its Second Vehicle in Ohio at Foxconn’s Newly Acquired Factory
- StoreDot Successfully Demonstrates its Rapid EV Battery Charging Technology That Could Deliver 100 Miles of Range in 5 Minutes
- General Motors and Red Hat to Collaborate on the Future of Software-Defined Vehicles
- Researchers at the Gwangju Institute of Science and Technology in South Korea Say They’ve Uncovered the Key to Safer EV Batteries
- Autonomous Truck Developer Embark Trucks Adds Freight Carrier U.S. Xpress to its Partner Development Program