ApolloScape | 2018 | - 140,000 annotated images
- 20,000 Lidar 3D point cloud annotation data
- No radar
| - Include GPS/IMU data/timestamps
- Different times in the day
- Mainly in Beijing, China
| - Semantic
- Instance-wise
- Dense pixel annotations
- 2D/3D boxes
|
BDD100K | 2017 | 100,000 annotated images (1280 * 720) | - Multiple cities
- Multiple scene type
- Different times in the day
- Include GPS/IMU data/timestamps
- Multiple weather
| - Semantic
- Instance-wise
- Dense pixel annotations
- 2D boxes
|
Cityscapes | 2016 | images (1280 * 720) | - 50 cities
- Several months
- Daytime
- Good weather conditions
| - Semantic
- Instance-wise
- Dense pixel annotations
|
| | | - Include GPS/timestamp metadata
| |
KITTI | 2012 | images (1248 * 384) - 15,000 Lidar 3D point cloud annotation data
| - Include GPS/IMU data/timestamps
- Daytime
| - Semantic
- Instance-wise
- Dense pixel annotations
|
Lyft Dataset | 2019 | - 55,000 3D annotated images
| - 1,000 driving scenes in multiple cities
- Different times in the day
| - Semantic
- Instance-wise
- Dense pixel annotations
|
nuScenes | 2019 | annotated images - 390,000 Lidar 3D point cloud data
| - 1,000 driving scenes in multiple cities
- Different times in the day
| - Semantic
- Instance-wise
- Dense pixel annotations
|
Waymo Open Dataset | 2019 | images (1920 * 1280 & 1920 * 886) - Lidar 3D point cloud data
- 12 million 3D labels and
| - 1,000 driving scenes in multiple cities
- Different times in the day
- Multiple weather (day and night, dawn and dusk, sun and rain)
| - Semantic
- Instance-wise
- Dense pixel annotations
|