High Resolution Sparse Voxel DAGs

This implementation is based on Github page library for handling octree IO operations.

My implementation files of SVDAG constructor from SVO are in the folder src/svdag_construction:

• SVDAG_main.cpp
• SVDAG.h
• SVDAG.cpp
• SVDAG_node.cpp
• SVDAG_node.h

Dependencies

The libraries needed are included in src/libs:

• libmorton
• libtri
• trimesh2
• glm

They are already set up and were built for Linux_x64 system. You could rebuild it using cmake

Building

After the libraries are build, the building scripts should be executed.

To rebuild the tri_convert, svo_constructor and svdag_constructor executables in the 'building_scripts' folder:

• ./build_tri_convert.sh
• ./build_svo.sh
• ./build_svdag.sh

If all the building scripts were executed successfully, the system is set up and could be used.

Execution pipeline

• tri_convert: A tool to convert any model file to a simple, streamable .tri format.
• svo_constructor: Out-Of-Core SVO Constructor: Partitioning, voxelizing and SVO Building rolled into one executable, needs a .tri file as input and return .octree, .octreenodes, .octreedata files as output.
• svdag_constructor: My implementation of the DAG reducing algorithm, needs the previous .octree, .octreenodes, .octreedata files and return a .dag file as output.

Execution steps: Step 1: Create .tri file from .ply file ./tri_convert -f ../models/bunny.ply

Step 2: Convert tri to octree file ./svo_constructor -f ../models/bunny.tri

Step 3: Create DAG from Octree ./svdag_constructor -f ../models/bunny1024_1.octree