search

Lishunkai / CNN-SLAM

The combination of CNN and SLAM, aiming to achieve a better result of traditional geometric-based SLAM architecture.
110 stars 38 forks source link

VIORB加入语义信息 #2

Open B-Betty opened 6 years ago

B-Betty commented 6 years ago

请问一下，VIORB加入语义信息可行吗？Visual-Inertial Monocular SLAM with Map Reuse，这篇文章是在orb-slam的基础上上融合IMU数据进行建图。

Lishunkai commented 6 years ago

Hello,

I believe that VIORB you mentioned refers to the newly proposed algorithm of the author of ORB-SLAM. In that way, I think it is impossible or hard to add semantic information of VIORB. Because ORB-SLAM is a typical feature-based SLAM, it produces sparse point clouds. The sparse point clouds is impossible (or hard) to add semantic information. Because if you want to implement semantic segmentation, you have to read RGB data as input, therefore it requires dense point clouds (map).

However, you can also use the method I mentioned to add semantic information to sparse point clouds, but I think it has no much use.

michaelczhou commented 6 years ago

Hello，既然单/双目相机无法提供稠密点云，那VIORB融合多线激光，可否做语义slam？ @Lishunkai