Scalable Rendering for Graphics and Game Engines (SRGGE) Lab Project

Table of contents

• General info
• How to run
• Implemented techniques
• Application output

General info

The goal of this project is to get familiar with various scalable techniques that allow us to render complex meshes.

How to run

Dependencies:

- (!) Important: Use in Linux. Developed and tested in Ubuntu, I don't know about other distributions.
- (!) Important: Use QtCreator 5. Different versions of Qt will probably cause errors.
- Eigen3 (to get: sudo apt install libeigen3-dev).

To compile:

- Go to BaseCode folder, and open a terminal there.
- run: mkdir build
- run: cd build
- run: cmake ..
- run: make

To run:

- First, run the Visibility executable. Inside the ~/build folder, run: ./Visibility
- This will create 2 visibility .txt files inside (origin of project)/extras/grid folder.
- Then, run the BaseCode executable. Inside the ~/build folder, run: ./BaseCode
- It will take some time to load the scene, because it will need to create the models' Levels-of-Detail (LODs) first.
- The created LODs will be stored in the same folder as the models inside (origin of project)/extras/models folder.
(!) Important: In case the loaded scene has "broken" meshes and LODs, terminate and re-run the BaseCode application.

Implemented techniques

• Created an ImGui for easy parameter configuration and method selection.
• Rendered a "complex" museum floor-wall plan, with 10 rooms full of meshes (+ central area), via a .txt translation.
• Computed simplified versions of the loaded model using vertex clustering on a regular grid, using a mean representative.
  • Extra: Used an octree to generate all simplified models simultaneously.
• Created a 2nd application (Visibility) to precompute cell-to-cell visibility using random visibility sampling (Simplified Bresenham).
  • Extra: Also implemented an optimization to the visibility precomputation (Supercover Bresenham).

Application output

Overview of the scene using simplified models