search

PRBonn / semantic_suma

SuMa++: Efficient LiDAR-based Semantic SLAM (Chen et al IROS 2019)
MIT License
923 stars 205 forks source link

different result with Disks mode #53

Closed aprilliuwei closed 2 years ago

aprilliuwei commented 2 years ago

Thanks for your code!This algorithm is very effective for dynamic elimination problems,but i still encounter some problems. when i use the kitti dataset,the display is normal, when i use the 16 line data trained by myself,the point mode seems to be normal,but when i switch to disks model,it belcomes a little strange,and the point cloud of the map becomes very large.How can i solve this problem?Another problem is that there is a large jitter in the process of running with 16 line data,i don't know why this phenomenon occurs.Thank you. kitti: Screenshot from 2022-01-07 21-20-45 Here is my data: Point mode: Screenshot from 2021-12-31 10-04-45 Disks mode: Screenshot from 2021-12-30 22-48-58

Chen-Xieyuanli commented 2 years ago

Hey @aprilliuwei,

The results with the 16-beam lidar look interesting, and thanks for the sharing and feedback.

We have a specific parameter for the disk maximum radius. You may check the parameter here: https://github.com/PRBonn/semantic_suma/blob/ca2d275729bbaa9b87b784793c191a8714a46d16/config/default.xml#L51