#010 3D Face Modeling – Gentle Introduction to 3D rotations on a cube
Highlights: Hello and welcome. In this post, we will see examples and how we can rotate 3D objects. We will see how we can rotate a 3D cube around each axis, the \(x, y, z\) axes.
Definition of the cube
Now, we will do an example in Python by creating a 3D object which will consist of points with 3 dimensions. This 3D object will be a cube and we will rotate it around the 3 axes, and using a simple projection we will project it into a 2D image.
First, let us first import all the needed dependencies. We will do all the drawings using OpenCV and matplotlib. Due to the use of google collab, for this example, we will import cv2_imshow to analyze images inside of it.
import numpy as np
import cv2
import matplotlib.pyplot as plt
from google.colab.patches import cv2_imshowCode language: JavaScript (javascript)We also need to set up parameters that will be used for image setup and initial angles, which will be zero values and no rotations. Next to these parameters, the center of the screen coordinate is calculated which will be used as a reference where to draw our cube.
WHITE = (255, 255, 255)
RED = (0, 0, 255)
BLACK = (0, 0, 0)
WIDTH = 800
HEIGHT = 600
scale = 100
circle_pos = [WIDTH//2, HEIGHT//2]
angle_x = angle_y = angle_z = 0Code language: JavaScript (javascript)Now, let us create a 3D object, in this case, a 3D cube. This cube will be later rotated around the x, y, and Z axes.
points = np.array ([[-1,-1, 1],
[ 1,-1, 1],
[ 1, 1, 1],
[-1, 1, 1],
[-1,-1,-1],
[ 1,-1,-1],
[ 1, 1,-1],
[-1, 1,-1]])
fig = plt.figure(figsize=(6, 6))
ax = fig.add_subplot(projection='3d')
ax.scatter(points[:, 0], points[:, 1], points[:, 2], c='r', s=100)
ax.set_xlim([-2, 2])
ax.set_xlim([-2, 2])
ax.set_zlim([-2, 2])
plt.show()Code language: PHP (php)You can see our 3D cube scattered in the plot below (Figure 1).
Additional functions and variables
Next, we will define our projection matrix, and because this is just an explanatory example, it will take the form of a simple projection matrix. We just remove the third axis, the Z axis. Alternatively to remove this \(3^{rd}\) row, just set the values in the third row to zeros. If you want to learn more about projection matrices, refer to this post.
projection_matrix = np.matrix([
[1, 0, 0],
[0, 1, 0]
])We will also create a list that will hold X and Y coordinates after the projection of our cube vertices.
projected_points = [
[n, n] for n in range(len(points))
]Creating a function connect_points that will draw lines between the points, connecting them.
def connect_points(screen, i, j, points):
cv2.line(screen, (points[i][0], points[i][1]), (points[j][0], points[j][1]), BLACK, 1)
return screenCode language: CSS (css)Understanding the rotation
Now comes the interesting part the rotation. Let us first do a rotation for one single frame and after we will see an animation.
The first thing we need to do is create a blank canvas. This will be achieved using np.ones() and creating an image of size (800×600).
screen = np.ones((HEIGHT, WIDTH, 3), dtype=np.uint8) * 255
print(screen.shape)
cv2_imshow(screen)Code language: PHP (php)
To be able to rotate something in 3D, we need to choose around which axes we want the rotation to happen. We can rotate a 3D object around one axis, or multiple, as we wish. For each axis, a different rotation matrix is used. Here are the matrices for each of the axis:
$$ R_x(\theta) = \begin{bmatrix} 1 & 0 & 0\\ 0 & \cos{\theta} & -\sin{\theta} \\ 0 & \sin{\theta} & -\cos{\theta} \end{bmatrix},R_y (\theta) = \begin{bmatrix} \cos{\theta} & 0 & \sin{\theta}\\ 0 & 1 & 0 \\ -\sin{\theta} & 0 & \cos{\theta} \end{bmatrix},R_z(\theta) = \begin{bmatrix} \cos{\theta} & -\sin{\theta} & 0 \\ \sin{\theta} & -\cos{\theta} & 0 \\ 0 & 0 & 1\\ \end{bmatrix} $$
This is how we can implement these rotation matrices in code.
rotation_z = np.matrix([
[np.cos(angle_x), -np.sin(angle_x), 0],
[np.sin(angle_x), np.cos(angle_x), 0],
[ 0, 0, 1],
])
rotation_y = np.matrix([
[ np.cos(angle_y), 0, np.sin(angle_y)],
[ 0, 1, 0 ],
[-np.sin(angle_y), 0, np.cos(angle_y)],
])
rotation_x = np.matrix([
[1, 0, 0 ],
[0, np.cos(angle_z), -np.sin(angle_z)],
[0, np.sin(angle_z), np.cos(angle_z)],
])
print("Rotation matrix around x axis:")
print(rotation_x)
print("Rotation matrix around y axis:")
print(rotation_y)
print("Rotation matrix around z axis:")
print(rotation_z)Code language: PHP (php)Output: Rotation matrix around x axis: [[ 1. 0. 0.] [ 0. 1. -0.] [ 0. 0. 1.]] Rotation matrix around y axis: [[ 1. 0. 0.] [ 0. 1. 0.] [-0. 0. 1.]] Rotation matrix around z axis: [[ 1. -0. 0.] [ 0. 1. 0.] [ 0. 0. 1.]]
Process of rotating a 3D cube
As we have defined our rotation matrices, let us see how we can plot our points and how they currently look in 2D. For this, we will use the projection matrix that we previously created and also the points. We will use the dot product to project the points from one dimension to the other. After that, we will scale our points, just so that we see our points in the drawing, and also we will translate/shift the points to the center of the screen.
We will save the values into the list that we have previously created so that we know where to draw the lines to connect the points. The remaining parts of the code are simply the drawing of points and lines.
Note: because we are drawing on OpenCV we need to reset the canvas every time we draw on it
screen = np.ones((HEIGHT, WIDTH, 3), dtype=np.uint8) * 255
i = 0
for point in points:
projected2d = np.dot(projection_matrix, point.reshape((3, 1)))
x = int(projected2d[0][0] * scale) + circle_pos[0]
y = int(projected2d[1][0] * scale) + circle_pos[1]
projected_points[i] = [x, y]
screen = cv2.circle(screen, (x, y), 20, RED, -1)
i += 1
for p in range(4):
screen = connect_points(screen, p, (p+1) % 4, projected_points)
screen = connect_points(screen, p+4, ((p+1) % 4) + 4, projected_points)
screen = connect_points(screen, p, (p+4), projected_points)
cv2_imshow(screen)Below in Figure 3, we can see the output image, where we can see that only 4 sides are visible because the rest are behind these four.
The interesting thing happens when we multiply the points with the rotation matrices. For example, if we just add a little bit to the Z axis and calculate the rotation matrix, and after we multiply the points with the rotation matrices, we can see that our cube has rotated. We can now see all the vertices of our cube.
angle_x = angle_y = angle_z = 0
angle_x += 0.00
angle_y += 0.25
angle_z += 0.00
rotation_z = np.matrix([
[np.cos(angle_x), -np.sin(angle_x), 0],
[np.sin(angle_x), np.cos(angle_x), 0],
[ 0, 0, 1],
])
rotation_y = np.matrix([
[ np.cos(angle_y), 0, np.sin(angle_y)],
[ 0, 1, 0 ],
[-np.sin(angle_y), 0, np.cos(angle_y)],
])
rotation_x = np.matrix([
[1, 0, 0 ],
[0, np.cos(angle_z), -np.sin(angle_z)],
[0, np.sin(angle_z), np.cos(angle_z)],
])
screen = np.ones((HEIGHT, WIDTH, 3), dtype=np.uint8) * 255
i = 0
for point in points:
rotated3d = np.matmul(rotation_z, point.reshape((3, 1)))
rotated3d = np.matmul(rotation_y, rotated3d)
rotated3d = np.matmul(rotation_x, rotated3d)
projected2d = np.matmul(projection_matrix, rotated3d)
x = int(projected2d[0][0] * scale) + circle_pos[0]
y = int(projected2d[1][0] * scale) + circle_pos[1]
projected_points[i] = [x, y]
screen = cv2.circle(screen, (x, y), 20, RED, -1)
i += 1
for p in range(4):
screen = connect_points(screen, p, (p+1) % 4, projected_points)
screen = connect_points(screen, p+4, ((p+1) % 4) + 4, projected_points)
screen = connect_points(screen, p, (p+4), projected_points)
cv2_imshow(screen)
Let us now create some animations. The same as before, just now in a for a loop. We will create a video with a rotation around only the \(X\) axis. We will create a video of 5 seconds with an FPS of 30.
angle_x = angle_y = angle_z = 0
out = cv2.VideoWriter("test.mp4", cv2.VideoWriter_fourcc(*"mp4v"), 30, (WIDTH, HEIGHT))
for _ in range(5*30):
screen = np.ones((HEIGHT, WIDTH, 3), dtype=np.uint8) * 255
rotation_z = np.matrix([
[np.cos(angle_x), -np.sin(angle_x), 0],
[np.sin(angle_x), np.cos(angle_x), 0],
[0, 0, 1],
])
rotation_y = np.matrix([
[np.cos(angle_y), 0, np.sin(angle_y)],
[0, 1, 0],
[-np.sin(angle_y), 0, np.cos(angle_y)],
])
rotation_x = np.matrix([
[1, 0, 0],
[0, np.cos(angle_z), -np.sin(angle_z)],
[0, np.sin(angle_z), np.cos(angle_z)],
])
angle_x += 0.02
angle_y += 0.05
angle_z += 0.01
i = 0
for point in points:
rotated2d = np.matmul(rotation_z, point.reshape((3, 1)))
rotated2d = np.matmul(rotation_y, rotated2d)
rotated2d = np.matmul(rotation_x, rotated2d)
projected2d = np.matmul(projection_matrix, rotated2d)
x = int(projected2d[0][0] * scale) + circle_pos[0]
y = int(projected2d[1][0] * scale) + circle_pos[1]
projected_points[i] = [x, y]
screen = cv2.circle(screen, (x, y), 20, RED, -1)
i += 1
for p in range(4):
screen = connect_points(screen, p, (p+1) % 4, projected_points)
screen = connect_points(screen, p+4, ((p+1) % 4) + 4, projected_points)
screen = connect_points(screen, p, (p+4), projected_points)
out.write(screen)
out.release()Code language: JavaScript (javascript)Now, let’s create some more interesting rotations. We will create videos with rotations around all the axes and also combined rotations around multiple axes.
Summary
In this post, we covered some very interesting topics. We have seen how we can rotate a 3D object around each axis and how to create different animations. This is interesting because it is connected with the concept of 3D Face models that we have seen in previous posts. All of the rotations and projections we have seen here can be transferred to 3D face models.
