After several years of using rendering engines such as Pixar's RenderMan, Arnold or Mantra, I was tempted by the idea of exploring ray tracing in greater depth while deepening my knowledge of Python. Yes, python. A language by nature rather slow, and therefore far from being the preferred language for this exercise.
Challenge accepted, project launched !
Render time: 0.55 seconds
pip install numpy pillowYou can run this example:
import time
from pathlib import Path
from ray_tracer.application import render_image_pipeline
from ray_tracer.domain import Camera, DomeLight, PointLight, Scene3D
from ray_tracer.infrastructure import (
CheckeredSphere,
NumpyRenderer,
NumpyRGBColor,
Sphere3D,
Vector3D,
)
renderer = NumpyRenderer()
scene = Scene3D(
[
Sphere3D(
Vector3D(0.55, 0.5, 3),
0.5,
NumpyPhysicalShader(
Diffuse(Vector3D(1, 1, 1), 1.0),
Specular(Vector3D(0, 1, 1), 0.5, 0.5),
0.0,
0,
0.01,
0,
),
),
Sphere3D(
Vector3D(-0.55, 0.1, 1),
1,
NumpyPhysicalShader(
Diffuse(Path("sourceimages/2k_mars.jpg"), 1.0),
Specular(Vector3D(1, 1, 1), 1.0, 0.5),
0,
0.1,
0.1,
0.0,
),
),
Sphere3D(
Vector3D(0, -99999.5, 0),
99999,
NumpyPhysicalShader(
Diffuse(Vector3D(1, 1, 1), 1.0),
Specular(Vector3D(0, 1, 1), 0.5, 0.5),
0.1,
0.5,
0.0,
1.0,
),
),
],
[ # TODO: use multiple lights
PointLight(1.0, Vector3D(-5, 5, -10)),
# PointLight(Vector3D(-2, 1, 2)),
DomeLight(0.1, Vector3D(1, 1, 1)),
],
Camera(Vector3D(0, 0.2, -2), int(1920 / 2), int(1080 / 2)),
)
output_path = Path("render.png")
start_time = time.time()
render_image_pipeline(scene, output_path, renderer)
print("Took", time.time() - start_time)