def main():
try:
os.makedirs(args.output_directory)
except:
pass
# Set GPU device
rtx.set_device(args.gpu_device)
# Initialize colors
colors = []
for n in range(args.num_colors):
hue = n / args.num_colors
saturation = 1
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
colors.append((red, green, blue, 1))
screen_width = args.image_size
screen_height = args.image_size
# Setting up a raytracer
rt_args = rtx.RayTracingArguments()
rt_args.num_rays_per_pixel = 1024
rt_args.max_bounce = 3
rt_args.supersampling_enabled = args.anti_aliasing
rt_args.next_event_estimation_enabled = True
rt_args.ambient_light_intensity = 0.1
cuda_args = rtx.CUDAKernelLaunchArguments()
cuda_args.num_threads = 64
cuda_args.num_rays_per_thread = 32
renderer = rtx.Renderer()
render_buffer = np.zeros((screen_height, screen_width, 3),
dtype=np.float32)
archiver = Archiver(
directory=args.output_directory,
total_scenes=args.total_scenes,
num_scenes_per_file=min(args.num_scenes_per_file, args.total_scenes),
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=args.initial_file_number)
camera = rtx.PerspectiveCamera(fov_rad=math.pi / 3,
aspect_ratio=screen_width / screen_height)
for _ in tqdm(range(args.total_scenes)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_objects(scene,
colors,
max_num_objects=args.max_num_objects,
discrete_position=args.discrete_position,
rotate_object=args.rotate_object)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for _ in range(args.num_observations_per_scene):
# Sample camera position
rand_position_xz = np.random.uniform(-2.5, 2.5, size=2)
rand_lookat_xz = np.random.uniform(-6, 6, size=2)
camera_position = np.array(
[rand_position_xz[0], 1, rand_position_xz[1]])
look_at = np.array([rand_lookat_xz[0], 1, rand_lookat_xz[1]])
# Compute yaw and pitch
camera_direction = rand_position_xz - rand_lookat_xz
camera_direction = np.array(
[camera_direction[0], 0, camera_direction[1]])
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera.look_at(tuple(camera_position),
tuple(look_at),
up=(0, 1, 0))
renderer.render(scene, camera, rt_args, cuda_args, render_buffer)
# Convert to sRGB
image = np.power(np.clip(render_buffer, 0, 1), 1.0 / 2.2)
image = np.uint8(image * 255)
image = cv2.bilateralFilter(image, 3, 25, 25)
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
def main():
try:
os.makedirs(args.output_directory)
except:
pass
# Colors
colors = []
for n in range(args.num_colors):
hue = n / args.num_colors
saturation = 1
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
colors.append(np.array((red, green, blue, 1)))
renderer = OffscreenRenderer(viewport_width=args.image_size,
viewport_height=args.image_size)
archiver = Archiver(
directory=args.output_directory,
total_scenes=args.total_scenes,
num_scenes_per_file=min(args.num_scenes_per_file, args.total_scenes),
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=args.initial_file_number)
for scene_index in tqdm(range(args.total_scenes)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_objects(scene,
colors,
objects,
max_num_objects=args.max_num_objects,
discrete_position=args.discrete_position,
rotate_object=args.rotate_object)
camera_distance = 3
camera = PerspectiveCamera(yfov=math.pi / 4)
camera_node = Node(camera=camera, translation=np.array([0, 1, 1]))
scene.add_node(camera_node)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for observation_index in range(args.num_observations_per_scene):
# Sample camera position
rand_position_xz = np.random.uniform(-3, 3, size=2)
rand_lookat_xz = np.random.uniform(-6, 6, size=2)
camera_position = np.array(
[rand_position_xz[0], 1, rand_position_xz[1]])
# Compute yaw and pitch
camera_direction = rand_position_xz - rand_lookat_xz
camera_direction = np.array(
[camera_direction[0], 0, camera_direction[1]])
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera_node.rotation = genearte_camera_quaternion(yaw, pitch)
camera_node.translation = camera_position
# Rendering
flags = RenderFlags.SHADOWS_DIRECTIONAL
if args.anti_aliasing:
flags |= RenderFlags.ANTI_ALIASING
image = renderer.render(scene, flags=flags)[0]
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
renderer.delete()
def main():
# Colors
colors = []
for n in range(args.num_colors):
hue = n / args.num_colors
saturation = 1
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
colors.append(np.array((red, green, blue, 1)))
floor_textures = [
"../textures/lg_floor_d.tga",
"../textures/lg_style_01_floor_blue_d.tga",
"../textures/lg_style_01_floor_orange_bright_d.tga",
]
wall_textures = [
"../textures/lg_style_01_wall_cerise_d.tga",
"../textures/lg_style_01_wall_green_bright_d.tga",
"../textures/lg_style_01_wall_red_bright_d.tga",
"../textures/lg_style_02_wall_yellow_d.tga",
"../textures/lg_style_03_wall_orange_bright_d.tga",
]
objects = [
pyrender.objects.Capsule,
pyrender.objects.Cylinder,
pyrender.objects.Icosahedron,
pyrender.objects.Box,
pyrender.objects.Sphere,
]
renderer = OffscreenRenderer(viewport_width=args.image_size,
viewport_height=args.image_size)
plt.tight_layout()
fig = plt.figure(figsize=(6, 3))
axis_perspective = fig.add_subplot(1, 2, 1)
axis_orthogonal = fig.add_subplot(1, 2, 2)
ims = []
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_objects(scene,
colors,
objects,
min_num_objects=args.num_objects,
max_num_objects=args.num_objects,
discrete_position=args.discrete_position,
rotate_object=args.rotate_object)
camera_distance = 5
perspective_camera = PerspectiveCamera(yfov=math.pi / 4)
perspective_camera_node = Node(camera=perspective_camera,
translation=np.array([0, 1, 1]))
orthographic_camera = OrthographicCamera(xmag=3, ymag=3)
orthographic_camera_node = Node(camera=orthographic_camera)
rad_step = math.pi / 36
total_frames = int(math.pi * 2 / rad_step)
current_rad = 0
for _ in range(total_frames):
scene.add_node(perspective_camera_node)
# Perspective camera
camera_xz = camera_distance * np.array(
(math.sin(current_rad), math.cos(current_rad)))
# Compute yaw and pitch
camera_direction = np.array([camera_xz[0], 0, camera_xz[1]])
yaw, pitch = compute_yaw_and_pitch(camera_direction)
perspective_camera_node.rotation = genearte_camera_quaternion(
yaw, pitch)
camera_position = np.array([camera_xz[0], 1, camera_xz[1]])
perspective_camera_node.translation = camera_position
# Rendering
flags = RenderFlags.SHADOWS_DIRECTIONAL
if args.anti_aliasing:
flags |= RenderFlags.ANTI_ALIASING
image = renderer.render(scene, flags=flags)[0]
im1 = axis_perspective.imshow(image,
interpolation="none",
animated=True)
scene.remove_node(perspective_camera_node)
# Orthographic camera
scene.add_node(orthographic_camera_node)
camera_direction = camera_distance * np.array(
(math.sin(current_rad), math.sin(
math.pi / 6), math.cos(current_rad)))
yaw, pitch = compute_yaw_and_pitch(camera_direction)
orthographic_camera_node.rotation = genearte_camera_quaternion(
yaw, pitch)
orthographic_camera_node.translation = np.array(
[camera_direction[0], 4, camera_direction[2]])
image = renderer.render(scene, flags=flags)[0]
im2 = axis_orthogonal.imshow(image,
interpolation="none",
animated=True)
ims.append([im1, im2])
plt.pause(1e-8)
current_rad += rad_step
scene.remove_node(orthographic_camera_node)
ani = animation.ArtistAnimation(fig,
ims,
interval=1 / 24,
blit=True,
repeat_delay=0)
filename = "rooms"
if args.discrete_position:
filename += "_discrete_position"
if args.rotate_object:
filename += "_rotate_object"
if args.fix_light_position:
filename += "_fix_light_position"
filename += ".gif"
ani.save(filename, writer="imagemagick")
def main():
# Set GPU device
rtx.set_device(args.gpu_device)
# Initialize colors
colors = []
for n in range(args.num_colors):
hue = n / args.num_colors
saturation = 1
lightness = 1
red, green, blue = colorsys.hsv_to_rgb(hue, saturation, lightness)
colors.append((red, green, blue, 1))
screen_width = args.image_size
screen_height = args.image_size
# Setting up a raytracer
rt_args = rtx.RayTracingArguments()
rt_args.num_rays_per_pixel = 1024
rt_args.max_bounce = 3
rt_args.supersampling_enabled = args.anti_aliasing
rt_args.next_event_estimation_enabled = True
rt_args.ambient_light_intensity = 0.1
cuda_args = rtx.CUDAKernelLaunchArguments()
cuda_args.num_threads = 64
cuda_args.num_rays_per_thread = 32
renderer = rtx.Renderer()
render_buffer = np.zeros((screen_height, screen_width, 3),
dtype=np.float32)
perspective_camera = rtx.PerspectiveCamera(fov_rad=math.pi / 3,
aspect_ratio=screen_width /
screen_height)
orthographic_camera = rtx.OrthographicCamera()
camera_distance = 2
plt.tight_layout()
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_objects(scene,
colors,
min_num_objects=args.num_objects,
max_num_objects=args.num_objects,
discrete_position=args.discrete_position,
rotate_object=args.rotate_object)
current_rad = 0
rad_step = math.pi / 36
total_frames = int(math.pi * 2 / rad_step)
fig = plt.figure(figsize=(6, 3))
axis_perspective = fig.add_subplot(1, 2, 1)
axis_orthogonal = fig.add_subplot(1, 2, 2)
ims = []
for _ in range(total_frames):
# Perspective camera
camera_position = (camera_distance * math.cos(current_rad),
wall_height / 2,
camera_distance * math.sin(current_rad))
center = (0, wall_height / 2, 0)
perspective_camera.look_at(camera_position, center, up=(0, 1, 0))
renderer.render(scene, perspective_camera, rt_args, cuda_args,
render_buffer)
image = np.power(np.clip(render_buffer, 0, 1), 1.0 / 2.2)
image = np.uint8(image * 255)
image = cv2.bilateralFilter(image, 3, 25, 25)
im1 = axis_perspective.imshow(image,
interpolation="none",
animated=True)
# Orthographic camera
offset_y = 1
camera_position = (2 * math.cos(current_rad),
2 * math.sin(math.pi / 6) + offset_y,
2 * math.sin(current_rad))
center = (0, offset_y, 0)
orthographic_camera.look_at(camera_position, center, up=(0, 1, 0))
renderer.render(scene, orthographic_camera, rt_args, cuda_args,
render_buffer)
image = np.power(np.clip(render_buffer, 0, 1), 1.0 / 2.2)
image = np.uint8(image * 255)
image = cv2.bilateralFilter(image, 3, 25, 25)
im2 = axis_orthogonal.imshow(image,
interpolation="none",
animated=True)
ims.append([im1, im2])
# plt.pause(1e-8)
current_rad += rad_step
ani = animation.ArtistAnimation(fig,
ims,
interval=1 / 24,
blit=True,
repeat_delay=0)
filename = "rooms"
if args.discrete_position:
filename += "_discrete_position"
if args.rotate_object:
filename += "_rotate_object"
if args.fix_light_position:
filename += "_fix_light_position"
filename += ".gif"
ani.save(filename, writer="imagemagick")
def main():
try:
os.makedirs(args.output_directory)
except:
pass
last_file_number = args.initial_file_number + args.total_scenes // args.num_scenes_per_file - 1
initial_file_number = args.initial_file_number
if os.path.isdir(args.output_directory):
files = os.listdir(args.output_directory)
for name in files:
number = int(name.replace(".h5", ""))
if number > last_file_number:
continue
if number < args.initial_file_number:
continue
if number < initial_file_number:
continue
initial_file_number = number + 1
total_scenes_to_render = args.total_scenes - args.num_scenes_per_file * (
initial_file_number - args.initial_file_number)
assert args.num_scenes_per_file <= total_scenes_to_render
# Load MNIST images
mnist_images = load_mnist_images()
# Set GPU device
rtx.set_device(args.gpu_device)
screen_width = args.image_size
screen_height = args.image_size
# Setting up a raytracer
rt_args = rtx.RayTracingArguments()
rt_args.num_rays_per_pixel = 1024
rt_args.max_bounce = 3
rt_args.supersampling_enabled = args.anti_aliasing
rt_args.next_event_estimation_enabled = True
rt_args.ambient_light_intensity = 0.1
cuda_args = rtx.CUDAKernelLaunchArguments()
cuda_args.num_threads = 64
cuda_args.num_rays_per_thread = 32
renderer = rtx.Renderer()
render_buffer = np.zeros((screen_height, screen_width, 3),
dtype=np.float32)
archiver = Archiver(
directory=args.output_directory,
num_scenes_per_file=args.num_scenes_per_file,
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=initial_file_number)
camera = rtx.PerspectiveCamera(fov_rad=math.pi / 3,
aspect_ratio=screen_width / screen_height)
camera_distance = 2
for _ in tqdm(range(total_scenes_to_render)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_dice(scene,
mnist_images,
discrete_position=args.discrete_position,
rotate_dice=args.rotate_dice)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for _ in range(args.num_observations_per_scene):
# Sample camera position
rand_position_xz = np.random.normal(size=2)
rand_position_xz = camera_distance * rand_position_xz / np.linalg.norm(
rand_position_xz)
camera_position = np.array(
(rand_position_xz[0], wall_height / 2, rand_position_xz[1]))
center = np.array((0, wall_height / 2, 0))
# Compute yaw and pitch
camera_direction = camera_position - center
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera.look_at(tuple(camera_position), tuple(center), up=(0, 1, 0))
renderer.render(scene, camera, rt_args, cuda_args, render_buffer)
# Convert to sRGB
image = np.power(np.clip(render_buffer, 0, 1), 1.0 / 2.2)
image = np.uint8(image * 255)
image = cv2.bilateralFilter(image, 3, 25, 25)
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
def main():
try:
os.makedirs(args.output_directory)
except:
pass
# Load MNIST images
mnist_images = load_mnist_images()
renderer = OffscreenRenderer(viewport_width=args.image_size,
viewport_height=args.image_size)
archiver = Archiver(
directory=args.output_directory,
total_scenes=args.total_scenes,
num_scenes_per_file=min(args.num_scenes_per_file, args.total_scenes),
image_size=(args.image_size, args.image_size),
num_observations_per_scene=args.num_observations_per_scene,
initial_file_number=args.initial_file_number)
for scene_index in tqdm(range(args.total_scenes)):
scene = build_scene(floor_textures,
wall_textures,
fix_light_position=args.fix_light_position)
place_dice(scene,
mnist_images,
discrete_position=args.discrete_position,
rotate_dice=args.rotate_dice)
camera_distance = 4
camera = PerspectiveCamera(yfov=math.pi / 4)
camera_node = Node(camera=camera, translation=np.array([0, 1, 1]))
scene.add_node(camera_node)
scene_data = SceneData((args.image_size, args.image_size),
args.num_observations_per_scene)
for observation_index in range(args.num_observations_per_scene):
rand_position_xz = np.random.normal(size=2)
rand_position_xz = camera_distance * rand_position_xz / np.linalg.norm(
rand_position_xz)
# Compute yaw and pitch
camera_direction = np.array(
[rand_position_xz[0], 0, rand_position_xz[1]])
yaw, pitch = compute_yaw_and_pitch(camera_direction)
camera_node.rotation = genearte_camera_quaternion(yaw, pitch)
camera_position = np.array(
[rand_position_xz[0], 1, rand_position_xz[1]])
camera_node.translation = camera_position
# Rendering
flags = RenderFlags.SHADOWS_DIRECTIONAL
if args.anti_aliasing:
flags |= RenderFlags.ANTI_ALIASING
image = renderer.render(scene, flags=flags)[0]
scene_data.add(image, camera_position, math.cos(yaw),
math.sin(yaw), math.cos(pitch), math.sin(pitch))
if args.visualize:
plt.clf()
plt.imshow(image)
plt.pause(1e-10)
archiver.add(scene_data)
renderer.delete()