class: center, middle # Creación de Videojuegos ### Homogeneous Coordinates, Matrices, Quaternions --- class: center, middle # Coordinates ## Or: Where do we draw things? --- # Coordinates $$ \vec{v} = \begin{pmatrix} x \\\\ y \end{pmatrix} $$ -- ## Movement $$ \vec{v}' = \vec{v} + \begin{pmatrix} \Delta x \\\\ \Delta y \end{pmatrix} $$ -- ## Rotation? Scaling? Combinations? --- # Homogeneous coordinates $$ \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} $$ --- class: small # Homogeneous coordinates $$ \begin{pmatrix} a_1 & b_1 & c_1 \\\\ a_2 & b_2 & c_2 \\\\ a_3 & b_3 & c_3 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} a_1\cdot x + b_1 \cdot y + c_1 \\\\ a_2\cdot x + b_2 \cdot y + c_2 \\\\ a_2 \cdot x + b_3 \cdot y + c_3 \end{pmatrix} $$ --- class: small # Homogeneous coordinates $$ \begin{pmatrix} a_1 & b_1 & c_1 \\\\ a_2 & b_2 & c_2 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} a_1\cdot x + b_1 \cdot y + c_1 \\\\ a_2\cdot x + b_2 \cdot y + c_2 \\\\ 1 \end{pmatrix} $$ --- class: small # Homogeneous coordinates $$ \begin{pmatrix} 1 & 0 & 2 \\\\ 0 & 1 & 0 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} 1\cdot x + 0 \cdot y + 2 \\\\ 0\cdot x + 1 \cdot y + 0 \\\\ 0 \cdot x + 0 \cdot y + 1 \end{pmatrix} $$ --- class: small # Homogeneous coordinates $$ \begin{pmatrix} 1 & 0 & 2 \\\\ 0 & 1 & 0 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} x + 2 \\\\ y \\\\ 1 \end{pmatrix} $$ --- class: small # Homogeneous coordinates $$ \begin{pmatrix} 1 & 0 & 0 \\\\ 0 & 1 & 2 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} 1\cdot x + 0 \cdot y + 0 \\\\ 0\cdot x + 1 \cdot y + 2 \\\\ 0 \cdot x + 0 \cdot y + 1 \end{pmatrix} $$ --- class: small # Homogeneous coordinates $$ \begin{pmatrix} 1 & 0 & 0 \\\\ 0 & 1 & 2 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} x \\\\ y + 2 \\\\ 1 \end{pmatrix} $$ --- class: small # Homogeneous coordinates: Translation $$ \begin{pmatrix} 1 & 0 & \Delta x \\\\ 0 & 1 & \Delta y \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} x + \Delta x \\\\ y + \Delta y \\\\ 1 \end{pmatrix} $$ --- # Homogeneous coordinates: Rotation? Let's rotate $$ \vec{v} = \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} $$ by 90 degrees --- # Homogeneous coordinates: Rotation? $$ \begin{pmatrix} a_1 & b_1 & c_1 \\\\ a_2 & b_2 & c_2 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} -y \\\\ x \\\\ 1 \end{pmatrix} $$ --- class: small # Homogeneous coordinates: Rotation $$ \begin{pmatrix} 0 & -1 & 0 \\\\ 1 & 0 & 0 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} -y \\\\ x \\\\ 1 \end{pmatrix} $$ --- class: small # How do you rotate by an arbitrary angle? How do you rotate $$ \vec{v} = \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} $$ by `\(\alpha \)`? --- class: small # How do you rotate by an arbitrary angle? What's the angle `\( \beta \)` between `\( \vec{v} \)` and the x axis? --- class: small # How do you rotate by an arbitrary angle? What's the angle `\( \beta \)` between `\( \vec{v} \)` and the x axis? $$ \tan(\beta) = \frac{y}{x} = \frac{\sin(\beta)}{\cos(\beta)} $$ Note: $$ x = c \cdot \cos(\beta) $$ $$ y = c \cdot \sin(\beta) $$ --- class: small # Rotate Now we rotate by `\(\alpha \)` $$ \tan(\beta + \alpha) = \frac{\sin(\beta + \alpha)}{\cos(\beta + \alpha)} = \frac{y'}{x'} $$ We want to construct a matrix such that $$ \begin{pmatrix} a_1 & b_1 & c_1 \\\\ a_2 & b_2 & c_2 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} x' \\\\ y' \\\\ 1 \end{pmatrix} $$ Recall:
`\(x = c \cdot \cos(\beta)\)`
`\(y = c \cdot \sin(\beta)\)` --- # What is the rotation supposed to do? $$ \begin{pmatrix} a_1 & b_1 & c_1 \\\\ a_2 & b_2 & c_2 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} c \cdot \cos(\beta) \\\\ c \cdot \sin(\beta) \\\\ 1 \end{pmatrix} = \begin{pmatrix} x' \\\\ y' \\\\ 1 \end{pmatrix} $$ $$ x' = a_1 \cdot c \cdot \cos(\beta) + b_1 \cdot c \cdot \sin(\beta) + c_1 $$ $$ y' = a_2 \cdot c \cdot \cos(\beta) + b_2 \cdot c \cdot \sin(\beta) + c_2 $$ $$ \frac{y'}{x'} = \frac{\sin(\beta + \alpha)}{\cos(\beta + \alpha)} = \frac{\sin(\beta) \cos(\alpha) + \cos(\beta) \sin(\alpha)}{\cos(\beta) \cos(\alpha) - \sin(\beta) \sin(\alpha)} $$ --- # We can choose the values $$ \frac{y'}{x'} = \frac{\sin(\beta) \cos(\alpha) + \cos(\beta) \sin(\alpha)}{\cos(\beta) \cos(\alpha) - \sin(\beta) \sin(\alpha)} $$ $$ a_1 \cdot c \cdot \cos(\beta) + b_1 \cdot c \cdot \sin(\beta) + c_1 = \\\\ d \cdot (\cos(\beta) \cos(\alpha) - \sin(\beta) \sin(\alpha)) $$ $$ a_2 \cdot c \cdot \cos(\beta) + b_2 \cdot c \cdot \sin(\beta) + c_2 = \\\\ d \cdot (\sin(\beta) \cos(\alpha) + \cos(\beta) \sin(\alpha)) $$ --- # Some algebra $$ c \cdot(\cos(\beta) \cdot a_1 + \sin(\beta)\cdot b_1) = \\\\ d \cdot (\cos(\beta) \cos(\alpha) - \sin(\beta) \sin(\alpha)) $$ $$ c \cdot (\sin(\beta) \cdot b_2 + \cos(\beta) \cdot a_2) = \\\\ d \cdot (\sin(\beta) \cos(\alpha) + \cos(\beta) \sin(\alpha)) $$ $$ R = \begin{pmatrix} \cos(\alpha) & -\sin(\alpha) & 0 \\\\ \sin(\alpha) & \cos(\alpha) & 0 \\\\ 0 & 0 & 1 \end{pmatrix} $$ --- # What's the point? - Yay, we got a matrix! - Why do we care? - Also, what were c and d? - Why would I ever use this? --- # Matrix multiplication! Matrix multiplication is associative, that means: $$ A \cdot (B \cdot \vec{v}) = (A \cdot B) \cdot \vec{v} $$ This means, we can compose rotations and translations, and then apply the
composed
transform to a vector. Especially useful if we have many vectors to manipulate! --- class: small # Neat tricks - (Our) matrices can be inverted - The inverse of a translation matrix performs the reverse translation - The inverse of a rotation matrix performs the reverse rotation - Rotation is always around the origin, what if our object is not at `\((0,0)\:\)` ? - Translate it to the origin first, then rotate, then translate it back to where it should be - `\( T \cdot R \cdot T^{-1}\)` --- class: small # More neat tricks - We can also do other transformations, like reflection, shears, scaling, etc. $$ \begin{pmatrix} 0.5 & 0 & 0 \\\\ 0 & 0.5 & 0 \\\\ 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ 1 \end{pmatrix} = \begin{pmatrix} 0.5x \\\\ 0.5y \\\\ 1 \end{pmatrix} $$ - The same approach can be used for 3D coordinates --- class: small # Homogeneous coordinates in 3D $$ \begin{pmatrix} x \\\\ y \\\\ z \\\\ 1 \end{pmatrix} $$ --- class: small # Homogeneous coordinates in 3D $$ \begin{pmatrix} a_1 & b_1 & c_1 & d_1 \\\\ a_2 & b_2 & c_2 & d_2 \\\\ a_3 & b_3 & c_3 & d_3 \\\\ 0 & 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ z \\\\ 1 \end{pmatrix} = \begin{pmatrix} a_1\cdot x + b_1 \cdot y + c_1 \cdot x + d_1\\\\ a_2\cdot x + b_2 \cdot y + c_2 \cdot y + d_2 \\\\ a_3 \cdot x + b_3 \cdot y + c_3 \cdot z + d_3 \\\\ 1 \end{pmatrix} $$ --- class: small # Translation in 3D
$$ \begin{pmatrix} 1 & 0 & 0 & \Delta x \\ 0 & 1 & 0 & \Delta y \\ 0 & 0 & 1 & \Delta z \\ 0 & 0 & 0 & 1 \end{pmatrix} \cdot \begin{pmatrix} x \\ y \\ z \\ 1 \end{pmatrix} = \begin{pmatrix} x + \Delta x \\ y + \Delta y \\ z + \Delta z \\ 1 \end{pmatrix} $$
--- class: middle, center # Rotation in 3D? ## Maybe we can reuse the 2D rotation? --- class: small # Rotation in 3D
$$ R = \begin{pmatrix} \cos(\alpha) & -\sin(\alpha) & 0 & 0 \\\\ \sin(\alpha) & \cos(\alpha) & 0 & 0 \\\\ 0 & 0 & 1 & 0 \\\\ 0 & 0 & 0 & 1 \end{pmatrix} $$
What does this do? -- * Apparently it leaves the z coordinate unchanged * Rotates x and y * Rotation around the z axis! --- class: small # Rotation around the y axis?
$$ R_y = \begin{pmatrix} \cos(\alpha) & 0 & \sin(\alpha) & 0 \\ 0 & 1 & 0 & 0 \\ -\sin(\alpha) & 0 & \cos(\alpha) & 0 \\ 0 & 0 & 0 & 1 \end{pmatrix} $$
Notice anything? --- class: small # Rotation around the x axis
$$ R_x = \begin{pmatrix} 1 & 0 & 0 & 0 \\ 0 & \cos(\alpha) & - \sin(\alpha) & 0 \\ 0 & \sin(\alpha) & \cos(\alpha) & 0 \\ 0 & 0 & 0 & 1 \end{pmatrix} $$
--- class: small # Rotation around arbitrary vector * What if we want to rotate around something other than the x, y or z axis? * Any rotation can be represented by these basis rotations! * Euler angles * We can compose rotations using matrix multiplication! * Order is important! $$ R = R_y \cdot R_x \cdot R_z $$ --- class: small # Practice What's the matrix that rotates an object by 60 degrees around the y axis, *after* it moves it `\(\sqrt{3}\:\:\)` units in the z direction? How does it change the vector v? $$ \vec{v} = \begin{pmatrix} 1 \\\\ 1 \\\\ 0 \end{pmatrix} $$ Helpful: $$ \sin(60) = \frac{\sqrt{3}}{2}, \cos(60) = \frac{1}{2}, \sqrt{3} \approx 1.73 $$ --- class: small # Gimbal Lock
--- # Interpolating between rotations
--- # Quaternions! * Alternative representation for 3D rotations * Use *four* coordinates/numbers * Can be converted to/from rotation matrices * Any rotation can be represented as a single rotation around an axis $$ q = \cos\left(\frac{\theta}{2}\right) + \sin\left(\frac{\theta}{2}\right) \cdot (x \cdot i + y \cdot j + z \cdot k) $$ --- # Quaternion properties Composition is easy: $$ q' = q \cdot p \not = p \cdot q $$ Self-inverse (for rotations): $$ q \equiv -q $$ Rotating a vector: $$ v' = q \cdot v \cdot q^{-1} $$ --- # Quaternions in Unity * Rotations in Unity *are* quaternions * You can multiply two quaternions to compose them * You can multiply a quaternion and a vector to rotate it * You can interpolate between quaternions to get smooth rotations: ``` Quaternion.Slerp(p, q, t) ``` --- class: middle, center # Rendering --- class: small # Recall from last week * Model coordinates need to be translated to world coordinates (where is the mesh) * World coordinates are then translated to Camera coordinates (where in the field of view of the camera is the mesh?) * The Camera coordinates need to be projected onto a Viewport (objects that are further away become smaller) --- class: small # Recall from last week * Model coordinates need to be translated to world coordinates (where is the mesh): *Model matrix* * World coordinates are then translated to Camera coordinates (where in the field of view of the camera is the mesh?): *View Matrix* * The Camera coordinates need to be projected onto a Viewport (objects that are further away become smaller): *Projection Matrix* --- class: small # View Matrix * So far we have talked about how to move the objects in our scene around * But where is the Camera? * That's where the *View Matrix* comes in Futurama, Season 2, Episode 14: *The engines don’t move the ship at all. The ship stays where it is and the engines move the universe around it.* --- # View Matrix
--- class: small # View Matrix * Instead of defining where the Camera is, we give it a fixed position * Then, we move the scene into the field of view of the Camera * The picture is always in the xy plane * Larger z coordinates means something is further away from the Camera! * Added bonus: We can do all these movements with a matrix --- class: small # Projection Matrix * Now we need to figure out what our objects look like * Objects that are further away should be smaller $$ \begin{pmatrix} 1 & 0 & 0 & 0 \\\\ 0 & 1 & 0 & 0 \\\\ 0 & 0 & 0 & 0 \\\\ 0 & 0 & 1 & 0 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ z \\\\ 1 \end{pmatrix} = \begin{pmatrix} x \\\\ y \\\\ 0 \\\\ z \end{pmatrix} $$ --- class: small # What now? * We said earlier that the last coordinate is always 1 * What are we supposed to do with this? $$ \begin{pmatrix} x \\\\ y \\\\ 0 \\\\ z \end{pmatrix} $$ --- class: small # Actually ... * To map from homogeneous coordinates to 3D vectors, we actually perform this operation: $$ \begin{pmatrix} x \\\\ y \\\\ z \\\\ w \end{pmatrix} \equiv \begin{pmatrix} x/w \\\\ y/w \\\\ z/w \end{pmatrix} $$ * Our w was just always 1, so we could just discard it --- class: small # Projection Matrix $$ \begin{pmatrix} 1 & 0 & 0 & 0 \\\\ 0 & 1 & 0 & 0 \\\\ 0 & 0 & 0 & 0 \\\\ 0 & 0 & 1 & 0 \end{pmatrix} \cdot \begin{pmatrix} x \\\\ y \\\\ z \\\\ 1 \end{pmatrix} = \begin{pmatrix} x \\\\ y \\\\ 0 \\\\ z \end{pmatrix} \equiv \begin{pmatrix} x/z \\\\ y/z \\\\ 0 \end{pmatrix} $$ * The further away an object is, the smaller it is * We projected our objects onto the xy plane at z=0 --- # Matrices: summary * We can represent arbitrary 3D movements as matrices * The camera placement is represented by moving the objects to a fixed camera * Projecting the objects onto the view plane is another matrix * All of these matrices are composed $$ M = M_P \cdot M_V \cdot M_M $$ --- class: center, middle # Let there be light --- # The Phong Reflection Model * The Phong Reflection Model is an empirical model for how lights are reflected off of surfaces * Consists of ambient, diffuse and specular components
--- # The Phong Reflection Model
R
V
N
L
$$ I = k_a \cdot i_a + \sum_m \left(k_d (L_m \cdot N) i_d + k_s (R_m \cdot V)^\alpha i_s \right) $$ --- class: small # How do we use this information? * Flat shading: Calculate one value for the center of each triangle, and color the entire triangle that way * Gouraud shading: Calculate one value for each vertex of a triangle and interpolate the color in between * Phong shading: Interpolate normals between the vertices of a triangle, and calculate the color for each pixel
--- # Note * Phong reflection *model*: Equation to determine how light affects the color of a point * Phong shading/Gouraud shading/Flat shading: Methods to *use* this equation * Shader: The *implementation*/*program* that performs shading --- # Alternatives: Non-Photorealistic Shading * Toon/Cel shaders!
--- class: center, middle # Animation --- # T-Pose
--- # 3D Character Models * The T-Pose exposes many parts of the model to be able to fill holes * Characters are therefore usually modeled this way * But we want the character to move! * How do people move? * Skeletons and muscles! --- # Skeletons
--- # Attaching the Skin
--- # Muscles * The animation clips are the "muscles" * They determine where each bone should go and when * The animator often only defines *keyframes* beTWEEN which the system *interpolates* * [Mixamo](https://www.mixamo.com/#/?page=1&query=gangnam&type=Motion%2CMotionPack) --- class: small # References *
OpenGL tutorial on matrices
*
Math for Game Programmers: Understanding Homogeneous Coordinates
*
Quaternion description video
*
Interactive videos about Quaternions
*
Unity Quaternion documention