Dataset teaches autonomous vehicles to understand traffic signs
Mapillary, a street-level imagery platform that uses computer vision to automate and scale mapping, launched what it is calling "the world’s largest and most diverse publicly available traffic sign recognition dataset to teach autonomous vehicles to understand traffic signs." Named the Mapillary Traffic Sign Dataset, the dataset consists of 100,000 images from around the world. The images reportedly feature high variability in everything from weather and times of day when the images have been taken to camera sensors and viewpoints. The company states that this is the first time such a large and diverse dataset has been launched for anyone to license to train their own traffic sign recognition systems.
Emil Dautovic, VP of Automotive at Mapillary, said, “Carmakers typically go out and get their own data to train their algorithms, but that means that they have low levels of variability in their training data. When it comes to teaching cars to see, more diverse input data means better results.”
Mapillary says that all 570 million images on the Mapillary platform have been uploaded by people and companies from all over the world. One hundred thousand of these images were selected for the Mapillary Traffic Sign Dataset. More than 300 different traffic sign classes have been verified and annotated, resulting in more than 320,000 labeled traffic signs across the images. Over 52,000 images have been fully verified and annotated by humans, with the remaining images annotated partially through Mapillary's computer vision technology.
This news comes months after research showed that inexpensive cameras have the potential to catch up with LiDAR in teaching autonomous vehicles to understand their surroundings, something that would reduce the cost of an autonomous vehicle by tens of thousands of dollars. Mapillary’s new dataset tackles a different part of the problem of perception, but Dautovic noted the diversity in the dataset is key in moving toward camera-based solutions.
Dautovic explained, “The strength in the Mapillary Traffic Sign Dataset really lies in the diversity of the input data. There are only a few other datasets on the market, and none of them has imagery from all over the world, simply because it would take too much effort for one, single player to get images from such a diverse set of locations on a global scale. We don’t actually need to do that at Mapillary, since all the images have been uploaded to the Mapillary platform by people across the entire world. With this dataset, we’re hoping to come closer to solving the problem of self-driving from cameras only, one step at a time.”
The Mapillary Traffic Sign Dataset is available on https://www.mapillary.com/dataset/trafficsign.