This dataset is an extended dataset based on CityUHK-X dataset.
Related paper: BEV-Net
The files can be loaded by using h5py.
| Image | Head Map | Feet Map | BEV Map | BEV Back-Projected to IV |
|---|---|---|---|---|
 |
 |
 |
 |
 |
- image
- bev_center: the corresponding world coordinate of the BEV map center
- bev_scale: the physical scale of the BEV map, unit: meter/pixel
- bev_map: the density map of people in the bird's eye view
- bev_coord: the coordinates of people in the BEV map
- head_map: the density map of people's head in the image view
- head_annotation: the annotation of people's head
- feet_map: the density map of people's feet in the image view
- feet_annotation: the annotation of the mid-point of people's feet
- world_coord: the world coordinates of people
- num_annotations: number of annotations in each image
- roi_mask: mask of region of interest
- camera_angle: camera pitch angle
- camera_height: camera height relative to the ground plane
- camera_fu: camera focal length (horizontal)
- camera_fv: camera focal length (vertical)