Closed utkutpcgl closed 2 years ago
@utkutpcgl yes, both the rider on the bicycle and the bicycle itself would be bounded with a single box with the class vehicle.bicycle
As an example:
import json
from nuscenes.nuscenes import NuScenes
nusc = NuScenes(version='v1.0-mini', dataroot='/data/sets/nuscenes', verbose=False)
sample_ann_token ='1796f2adb80543d7a90ac9b62e11f79d'
print(json.dumps(nusc.get('sample_annotation', sample_ann_token), indent=4))
nusc.render_annotation(sample_ann_token)
The above code snippet will give you:
{
"token": "1796f2adb80543d7a90ac9b62e11f79d",
"sample_token": "1c9a906c40f6417bbe1cea06d6e55501",
...
"visibility_token": "4",
"attribute_tokens": [
"a14936d865eb4216b396adae8cb3939c"
],
"translation": [
683.746,
1588.663,
0.89
],
"size": [
0.717,
1.864,
1.932
],
...
"num_lidar_pts": 106,
...
"category_name": "vehicle.bicycle"
}
Thank you very much, I was about to do the same :)
@utkutpcgl I just realized you were asking about nuImages (not nuScenes) above, but you can use the same concept I suggested above to check if a single bounding box encompasses both a bike and its rider when it comes to nuImages
If it helps to others, I noticed that there is another extra annotation for bicycles showing the rider state:
with_rider and without_rider annotations was what I was looking for.
Thanks for your great contribution!
I have a simple question about nuImages bounding box annotation.
I am aware when that when a person is on a motorcycle, the bounding box covers them both with the label "vehicle.motorcycle". I am confused if the same applies to
vehicle.bicycle
.So, when a person is on a bicycle does the bounding box cover both the person and bicycle while the label is
vehicle.bicycle
?Thanks in advance.