Hi,
I am currently trying to project the point cloud in the image, but to me it looks like the camera extrinsics are not accurate. Probably the pitch is off?
I noticed the issue on multiple scenes, usually its pretty clear on the traffic lights.
e.g.
(Also on the left side the traffic lights are off)
It is also visible in 0464 and there isnt even any ego motion.
Is that to be expected?
You can reproduce the issue with the the devkit as follows:
from nuscenes.nuscenes import NuScenes
nusc = NuScenes(version='v1.0-trainval', dataroot='/home/data/datasets/public/nuscenes/data/', verbose=False)
SCENE_NAME = "scene-0480" # or 0103
scene = [x for x in nusc.scene if x["name"] == SCENE_NAME][0]
samples = sorted([x for x in nusc.sample if x["scene_token"] == scene["token"]], key=lambda x: x["timestamp"])
FRAME_ID = 14 # or 17
sample = samples[FRAME_ID]
nusc.render_pointcloud_in_image(sample["token"], pointsensor_channel='LIDAR_TOP')
Hi, I am currently trying to project the point cloud in the image, but to me it looks like the camera extrinsics are not accurate. Probably the pitch is off? I noticed the issue on multiple scenes, usually its pretty clear on the traffic lights. e.g.
(Also on the left side the traffic lights are off)
It is also visible in 0464 and there isnt even any ego motion.
Is that to be expected?
You can reproduce the issue with the the devkit as follows: