Tool that takes a video and creates a 3d model from it

[calledit]

Tool that stitches together a 3D mesh baed on depth maps from various angles.

People generally use point clouds for this or KinectFusion for this. Point clouds work by give unsatisfactory results. Ans it seams like KinectFusion assumes that the depth map is more accurate than todays monocular depth maps.
Python KinectFusion

• https://github.com/shiyoung77/KinectFusion-python
• https://github.com/JingwenWang95/KinectFusion
• https://github.com/andyzeng/tsdf-fusion-python

This tool just tries to stitch the models as they are.

[calledit]

The basic functionality of the tool is done.
The one thing that is left is bridging the gap between the stitched meshes.

This would be done by walking along the edges if each stitch and connecting every two vertexes to the closest vertex on the other side of the edge, once for each side of the edge.

The big problem is tracking, the current tracking accuracy does not seam good enough to generate nice meshes. Working on improving megasam camera tracking by supplementing it with new tracking markers might solve this.

[calledit]

Tried to move over to open3ds Tensor version...
Note: don't try that it again. That version is not done, you cant render headless on OSX you cant render without shadows at all. Nah.