Hi everyone,

So I posted a while back about my final year project and got some really good info.

However I have since changed my mind. Originally I wanted to do some graphics and with gpgpu cloth simulation.

I now want to do something more focused towards graphics if possible.


I recently submitted a proposal to make an adcanced 3d game engine, but got turned down due to the complexity and because it didn't have an "academic underpinning".


Now I turn to the Game Dev community to try and find a suitable project.

Time scale: 24 weeks
Main focus: graphics
Must be suitable for academic purposes - i.e. there has to be a reason to do it


I primarilly code in c++ and have a very good understanding of OpenGL.

Can anyone suggest a suitable idea?

Thanks
Nick

My Bachelor's thesis was frustum culling. I implemented single-threaded brute force culling and optimized it using AVX instructions and cache. It goes over 400,000,000 culls / second.

For Master's thesis I had an idea for occlusion culling. Later I found out that such method already exists and it is called coverage buffer occlusion culling. I still went with it. Used Intel's Software Culling as a base and replaced depth buffer rasterization with coverage buffer. On a side note it was only 1% better, but average FPS has significantly lower deviation.

You could also try to go for various shading algorithms. For example try to create shader for realistic human skin or animal fur. Maybe give a try at area lights (https://labs.unity.com/article/real-time-polygonal-light-shading-linearly-transformed-cosines). Maybe beat a dead horse and create Yet Another Planet Renderer™, possibly with full vegetation and simulation to show how planet evolves as time goes on (plants growing, dying, spreading, river erosion, deserts growing / disappearing, etc.).

For academic, try Vulkan API. Maybe you can discover something that may be useful for us too with that new API. :p

My Bachelor's thesis was frustum culling. I implemented single-threaded brute force culling and optimized it using AVX instructions and cache. It goes over 400,000,000 culls / second.

For Master's thesis I had an idea for occlusion culling. Later I found out that such method already exists and it is called coverage buffer occlusion culling. I still went with it. Used Intel's Software Culling as a base and replaced depth buffer rasterization with coverage buffer. On a side note it was only 1% better, but average FPS has significantly lower deviation.

You could also try to go for various shading algorithms. For example try to create shader for realistic human skin or animal fur. Maybe give a try at area lights (https://labs.unity.com/article/real-time-polygonal-light-shading-linearly-transformed-cosines). Maybe beat a dead horse and create Yet Another Planet Renderer™, possibly with full vegetation and simulation to show how planet evolves as time goes on (plants growing, dying, spreading, river erosion, deserts growing / disappearing, etc.).

Hey, thanks for the awesome reply!

One thing, with the bachelors final year project what was the academic reasoning for doing the project?

Thanks