Try out various things to get the reconstruction to work right on a 50-frame sequence
Closed this issue · 0 comments
Algomorph commented
The fusion result so far is awfully distorted.
Here is a list of (some of) the things that can be attempted.
- Tweak TSDF parameters, such as voxel size & truncation distance
-
Add negative signed distance cutoff parameter, like mu in KinectFusion / delta in KillingFusion[ Deferred ] - Test with just using previous frame mode (vs. the rendered + previous frame overlay mode)
- Try shortest path distance when computing anchors + weights during forward-warping the mesh
- Try shortest path distance when computing anchors + weights for the point-cloud input to the tracker
- Try shortest path distance when computing anchors + weights during the actual fusion process (?)