-
Rasterization: a color-encoded image of all map?
Data-type: image -
Vectorization: topology
Data-type: table(csv)
--> Be divided into 2 type: static and dynamic-
Static: such as lane, crosswalk, ...
- id ( group by - type to get object ) - x ( point's position | absolute <br> - y of shape ) | value - status: additional information
-
Dynamic: such as vehicle, pedestrian, traffic light, ...
has property and state-
Property:
- id - type - width - length
-
State:
- timestamp - id - center_x ( center point | absolute | also be - center_y of object ) | value | ground-truth - heading (yaw - align with map orientation) - status: depend on object type __ light_state = {"RED", "YELLOW", "GREEN"} __ velocity for moving object __ ...
|--> Data will be saved as short scene (~20 seconds?)
-
-
- Prediction
|__ Scene1
| |__ Images
| | |__ 10000000.jpeg
| | |__ 10000001.jpeg
| | ...
| |__ Points
| |__ static.csv
| |__ dynamic_property.csv
| |__ dynamic_state.csv
|
|__ Scene2
...
- Current object type:
Static | Dynamic |
---|---|
waypoint | car |
l_lane | motorbike |
r_lane | traffic_light |
crosswalk | pedestrian |
[traffic_sign] | bicycle |
- [traffic_sign] could be any type of traffic sign, for example:
- speed_limit.30
- speed_limit.60
- speed_limit.90
- ...
- pedestrian and bicycle have not been supported yet.
- A lot of things to do...