Homework 4

Code:

https://github.com/CyberHolmes/CSCI5607/tree/main/HW4

Files which I created or modified:

• ModelLoad.cpp

• ModelLoad_dirLight_blinnPhong.cpp

• ModelLoad_dirLight_Gouraud.cpp

• ModelLoad_dirLight.cpp

• ModelLoad_pointlight.cpp

All files above has optional command line input option to use different model files, default ./models/teapot.txt, e.g.

$ ./ModelLoad_dirLight

$ ./ModelLoad_dirLight ./models/<FILENAME>.txt

Additionally ModelLoad_xx..xx.cpp has camera and keyboard control to move, rotate and translate camera view live, as well as object color change, object rotation and light rotation control.

User Controls

key 'a' - move the camera to the right

key 'd' - move the camera to the left

key 'w' - move the camera down

key 's' - move the camera up

Hold down mouse left button and move the mouse - rotate camera horizontal and vertical angle

Roll mouse wheel - increase/decrease camera FoV

key 'c' - change object color randomly

key 'space' - start/stop light rotation

key 'm' - start/stop object rotation

key 'r' - reset camera view and object color

key 'q' and escape - exit program

I forgot to demonstrate in the demo that moving the teapot using keys while rotating with mouse can turn the teapot

Section A

Section B

Q1. When calling SDL CreateWindow() we passed 100 for the second and third parameters. Try changing these numbers, what is the effect?

Ans: The second and the third parameters in SDL_CreateWindow() specifies the location of the top left corner of the window object. Changing these numbers results in different location of the window generated later in the code.

Q2. In Cube3D.cpp we call glEnable(GL DEPTH TEST). Remove this call, how do you explain the resulting image?

Ans: Removing this call causes the rendered image to have the wrong triangle in the front. The reason is that this call makes opengl perform test the depth value of a fragment against the value in the depth buffer. This test ensures that only the fragment in the front get rendered.

Q3. Currently, all of our examples exits when escape is pressed. Update the code to also exit when the “Q” key is pressed. What did you need to change?

Ans: Add following code snippet in the while (SDL_PollEvent(&windowEvent)) loop.

if (windowEvent.type == SDL_KEYUP && windowEvent.key.keysym.sym == SDLK_q)

quit = true;

Q4. Create a new model file, by hand, that contains a single large triangle. Load this model instead of teapot.txt. Upload a picture of the rendering of this large triangle.

Ans: I created ./models/MyTriangle.txt. Below videos shows the rendering result of this model using ModelLoad.cpp

Completed Extra Credit Items (Results are shown below in video demo and images)

Computer exercise 1.1

Add a key-board based camera controller to one of the 3D examples. One simple approach is to add new variables which get passed into the call of glm::lookAt(). These variables should then be modified based on key press events.

Computer exercise 1.2

Update ModelLoad.cpp to rotate and translate the model based on key presses.

Computer exercise 1.3

Replace the phong specularity model in ModelLoad.cpp with Blinn-Phong. What happens to the size of the highlight?

Ans: Size of the highlight increases

Computer exercise 1.4

The vertex shader in ModelLoad.cpp hard codes the color to green. Update this code to instead draw the model loaded with a random color. Each time the user presses the ’c’ key change the model to a new random color.

Computer exercise 1.5

The vertex shader in CubeLit.cpp and ModelLoad.cpp hard codes the light to be shining in the direction of (1,0,0). Change this value from a const vec3 to a uniform, then dynamically update the direction to rotate around the model.

Computer exercise 1.6

The shaders used in CubeLit.cpp and ModelLoad.cpp support a directional light model. This means the strength of the lighting depends only on the direction of the light, not it’s distance. Update the code to support a point light. This change involves two steps. One, the direction that was statically stored in inlight- Dir should now be dynamically computed to point from the pos of the part of the model that is being rendered to current position of the light. Two, the strength of the light should be divided by the distance form the light, squared (i.e., further way lights should have less impact).

Computer exercise 1.7

The code shown in CubeLit.cpp and ModelLoad.cpp computes the full Phong lighting model for each pixel. This leads to a nice image, but can be expensive to compute. Re-write the shaders to move the entire lighting computation to the vertex shader. This approach is known as Gouraud shading, and can be much faster as the lighting is only computed once per vertex. By default, the graphics card will quickly interpolate the computer color between vertices. Is the rendering quality noticeably different? Which is faster?

Ans: The rendering quality can be noticeably different on some objects, such as sphere and knot. It is less significant on teapot. Gouraud should be faster since the color model computation is done per vertex instead of per pixel. However, the benchmark did not show difference on my compueter with those simple models.

Keyboard And Mouse Control Demo (Camera Control, Color Control, Light Rotation, Object Rotation)

Phong Lighting Vs. Blinn-Phong Lighting

Blinn-Phong lighting results in a larger specular area than Phong lighting

Gouraud Lighting Interpolation Vs Phone Lighting Interpolation

Point Light Vs Direction Light

Writeup:

I found opengl to be a little hard to understand at first. It took me a while to understand which part is handled by opengl and which part should be done by the programmer. This homework is a great way for me to ease into the world of opengl and graphics pipeline.