The Mapillary Traffic Sign Dataset is the world’s largest and most diverse publicly available traffic sign dataset for teaching machines to detect and recognize traffic signs. The dataset consists of 100,000 images from all over the world, with high variability in everything from weather and time of day to camera sensors and viewpoints.
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, using our computer vision technology.
We have run extensive experiments to establish strong baselines for both the detection and the classification tasks. In addition, we have verified that the diversity of this dataset enables effective transfer learning for existing large-scale benchmark datasets on traffic sign detection and classification.
We have also studied the impact of transfer learning using our traffic sign dataset and other traffic sign datasets released in the past. Our results show that pretraining on our dataset boosts the average precision of the binary detection task by ~6%, thanks to the completeness and diversity of our dataset.
- 100,000 high-resolution images (52,000 fully annotated, 48,000 partially annotated)
- Over 300 traffic sign classes with bounding box annotations
- Global geographic reach of images and traffic sign classes, covering 6 continents
- Variety of weather, season, time of day, camera, and viewpoint