The semantic annotation dataset contains about 550,000 labeled targets within 50 fragments. All those signs and lanes on the road are labeled with semantical annotation. We annotated them with 11 categories, e.g fence, solid line, lane marking, and bikeway, etc. More fragments will be collected and annotated in the future to continuously update the dataset.
Category：
Here is the category table for our dataset:
| category | Number of appearance | Rate of appearance |
| Fence | 19580 | 3.56% |
| Road | 144890 | 26.36% |
| Background | 13200 | 2.40% |
| Traffic sign | 81840 | 14.88% |
| Lane line | 79200 | 14.40% |
| Sidewalk | 50765 | 9.23% |
| Lane marking | 28270 | 5.14% |
| Crosswalk and diversion line | 26070 | 4.74% |
| Traffic light | 32945 | 5.99% |
| Person | 39435 | 7.17% |
| Car | 33715 | 6.13% |
Download：
Note: If you want to download our dataset, please send us an email with information about you and your organization. We will be glad to send you the passcode for the cloud storage.
Tsinghua University
Haidian District, Beijing, 100084, P. R. China
Email:xyzhang@tsinghua.edu.cn
