Obtaining Dense Point Cloud from VSLAM

[darylchooo]

We are extracting a point cloud after running stella vslam and we are getting a really sparse point cloud because stella vslam uses keyframes. is there any way we can obtain more dense point clouds using all the frames?

[Inshu32]

@Skwangles @ymd-stella

[Skwangles]

Assuming you are asking re dense pointcloud reconstruction? I'd love to know if I'm wrong, but I'm don't think that's possible with the current system. SLAMs focus is for more autonomous robotic navigation and rough camera pose tracking.

Though I'm sure if you could figure out how dense point clouds reconstruction works, it could be shoehorned in to work with the guestimate landmark positions.

I'd suggest looking into 3DGrut by Nvidia, VisualSFM, OpenSFM, PixSFM.

Any particular reason you'd need vslam?

[Skwangles]

If you are alternatively asking for more landmark generation, you can experiment with some of the setting to increase the number of landmarks it stores and the number of key frames it makes - though likely to have a performance impact