Code:

• • https://github.com/CyberHolmes/CSCI5607/tree/main/Project3

Team Member:

• Hailin Archer

Description

This ray tracer program takes a scene text file from commandline as input. It renders the scene on computer screen using openGL. It also saves the image on disk as specified by the scene file. It allows user to control the camera's position and view angle by keyboard input.

How to compile the code

$ make clean

$ make

How to run the program

$ ./ray ./Scenes/<FILENAME>.txt

User Controls

'a' - control camera x axis by reducing x component

'd' - control camera x axis by adding x component

's' - control camera y axis by reducing x component

'w' - control camera y axis by adding x component

'e' - control camera z axis by reducing x component

'q' - control camera z axis by adding x component

'j' - control camera forward direction's x component by reducing x component

'l' - control camera forward direction's x component by adding x component

'k' - control camera forward direction's y component by reducing y component

'i' - control camera forward direction's y component by adding y component

'u' - control camera forward direction's z component by reducing z component

'o' - control camera forward direction's z component by adding z component

'b' - save current rendering result on screen to a file with time stamp (won't delete the image saved at the beginning of the rendering)

'escape' - quit program <implemented in starter code>

Images with shadow, adaptive super sampling, reflection, refraction, multiple lights, directional light.

Sampling Method Comparison

OpenGL and UI

Writeup

At the beginning of the project, I spent most of my time getting up to speed with intermediate to advanced c++ techniques. I started with a object oriented design approach and made a ULM class diagram. When I start implementing it, I found this approach is not practical in that it will take too much time in unit testing, debugging and modifying the design. Instead I expanded upon the file parser, sphere intersection test and image library from HW3 codes, adding additional classes. This approach is much less overwhelming and I was able to make progress rapidly. My final code design is mainly object oriented. There are two major classes which does most of the heavy lifting: objects and scene. The objects classes defines everything in the scene and it comes with the shading function which returns hit information or color. The scene class stores objects in the scene and does the ray tree recursion.

Recursive call, reflection and refraction did not take long to implement, since my code structure supports it very well.

I found sometimes it was the trivia mistake that cost me time. At the beginning of the project, when calculating ray collision position, I took the first valid root, instead of checking for which root is the closest. I was puzzled by the result for a long time and kept checking my ray casting implementation until I realized what is happening. When I implemented the spot light, I forgot to convert degree to radian. The result did not look right of course. I focused on more complicated vector calculation for bug and it took me a while to find out it was the input value issue. And Dr. Guy helped me find out that what caused my refraction not working right is that I forgot to apply square root in the equation.

Another realization from the project is how powerful compiler can be. Under Dr. Guy's suggestion, I compiled the code using -O3 optimization, instead of standard compilation. The optimized executable's run time is half of the un-optimized compilation.

Art Contest