search

nv-tlabs / lift-splat-shoot

Lift, Splat, Shoot: Encoding Images from Arbitrary Camera Rigs by Implicitly Unprojecting to 3D (ECCV 2020)
Other
1.06k stars 221 forks source link

there is no lane segmentation ? #27

Closed Kenneth-X closed 2 years ago

Kenneth-X commented 2 years ago

I check the nuscenes datasets, but i find there is no lane gt label : https://www.nuscenes.org/nuscenes#data-annotation so, i wonder how does it work to train for lane segmentation and modify the code image

gsg213 commented 2 years ago

Hi @Kenneth-X,

You are right there isn't a lane segmentation class, in order to find it, you need the nuScenes Map Expansion and extract the ground truth from NuScenesMap.

You can go to https://www.nuscenes.org/nuscenes?tutorial=maps and you will find a class called lane_divider.

Finally, you will need to adapt the data loader to bring the map section you require.

I hope this helps!