Skip to content

Leaderboard doesn't give Semantic Lidar data format properly as mention in Carla docs #190

Description

@RohitPawar2406

{ "type": "sensor.lidar.ray_cast_semantic", "x": 1.3, "y": 0.0, "z": 2.5, "roll": 0.0, "pitch": 0.0, "yaw": -90.0, "horizontal_fov": 360.0, "id": "lidar_sem", },

Above is Semantic Lidar attributes mentioned. While collecting data through leaderboard script I found that semantic lidar data has shape (N, 6) . So technically frist three values are x,y,z (floats) next is cosine values (float) and object_indx and semantic_tag should be uint8 BUT I AM GETTING FLOAT32.
Below are images where I applied breakpoint() after entering in run_step function.
Screenshot from 2025-01-14 01-43-03
Screenshot from 2025-01-14 01-44-43

I will also add issues from Carla repo: https://github.com/carla-simulator/carla/issues/8419 , After going this issue you will understand there is no need to make semantic lidar data completely float, last two columns (object hit indx and semantic tag indx ) must be uint8.
I think this might be issue previous on Carla simulator mentioned in https://github.com/carla-simulator/carla/issues/8367
But now they have solved when we write script of Carla client side script from sctrach and mention semantic lidar. As I have tested that also.

But my issue is Semantic Lidar data is not properly in datatype manner while using Leaderboard script. Please see into this thing. And let me know if I am wrong somewhere....
@ICGog @felipecode @cmpute @joel-mb

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions