Optimal Meshing Process-Meshroom experimentations 190610

WARNING. This is a workin progress article and sections are not completed.

@Decisions :

Using LSCM as Unwrapping

Draft are interestingly fast and acceptable for conceptualize

@Analysis

Generating 3 Texturing of two types. Draft and High-Res
(ref as : HMeshing and DMeshing)

The variant being evaluated is the Unwrapping method

LSCM
DMesh Texturing – One LSCM was generated

DMesh Texturing Basic Unwrapping
DMeshing Basic generated two plus other unknown files

Generating ABF Unwrapping… Interesting

ABF Unwrapping
DMesh ABF Unacceptable – Closer though the result are a bit blurry, lets try generating for the HMesh
DMesh texturing ABF Unwrapping
TexturesOne Texture
MeshSimilar size than others
24min…

Generating the ABF for the HMesh
Start mesh atlasing (using Geogram ABF).
atlasing : Interesting term… can you clarify atlasing?

Canceled the ABF, way too long…


In realtime computer graphics, a texture atlas (also called a sprite sheet or an image sprite) is an image containing a collection of smaller images, usually packed together to reduce the atlas size. Atlases can consist of uniformly-sized sub-images, or they can consist of images of varying dimensions.


https://en.wikipedia.org/wiki/Texture_atlas

More experimentation on the tree were done.

Feature Matching

@q What is the optimal parametering for Feature matching?
In this experimentation, SIFT + AKAZE + Guided Matching was used

Texturing…

In conclusion. The Texturing LSCM is the best I have found to use.

16k Texture, EXR format, Unwrapping using method LSCM
Later, the 16k can be reduce to 8k, 4k according to desired resolution.

Meshing

Lower the triangle reduce a bit the generated model
Keep only the largest Mesh, well should keep only the largest mesh and get rid of all junk around

SfM – Structure from Motion

Increased
minInputTrackLength set to 4 to keep only the most robust matches (ref)

Error Memory Exausted 🙁

It does not like High feature extraction for some of my models